Policies

Recipes

Catalog of regex rules, suppressions, sensitive tools, and judge categories shipped with the strict policy — search, copy, and remix.

The recipe catalog is the same library the Policy creator draws from. Every entry ships with the bundled strict policy so you can paste the YAML directly into your own pack — or use the wizard to drop one into a draft and tune it interactively.

Where these come from

Each rule originates in policies/guardrail/strict/rules/<category>.yaml and gets compiled into this catalog at build time by docs-site/scripts/build-policy-assets.ts. The bundled strict pack is the source of truth — when it changes, this catalog regenerates.

Showing 192 of 192 recipes.
  • webhook.site (known exfil)rule:c2
    RECIPE-C2-WEBHOOK-SITE
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-WEBHOOK-SITE",
      "pattern": "(?i)(?:^|[^a-zA-Z0-9-])webhook\\.site(?:[^a-zA-Z0-9.-]|$)",
      "title": "webhook.site (known exfil)",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • ngrok tunnel (exfil risk)rule:c2
    RECIPE-C2-NGROK
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-NGROK",
      "pattern": "(?i)(?:^|[^a-zA-Z0-9-])(?:ngrok\\.io|ngrok-free\\.app)(?:[^a-zA-Z0-9.-]|$)",
      "title": "ngrok tunnel (exfil risk)",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • Pipedream (known exfil)rule:c2
    RECIPE-C2-PIPEDREAM
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-PIPEDREAM",
      "pattern": "(?i)pipedream\\.net",
      "title": "Pipedream (known exfil)",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • RequestBin (known exfil)rule:c2
    RECIPE-C2-REQUESTBIN
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-REQUESTBIN",
      "pattern": "(?i)requestbin\\.com",
      "title": "RequestBin (known exfil)",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • HookBin (known exfil)rule:c2
    RECIPE-C2-HOOKBIN
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-HOOKBIN",
      "pattern": "(?i)hookbin\\.com",
      "title": "HookBin (known exfil)",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • Burp Collaborator (pentest C2)rule:c2
    RECIPE-C2-BURP
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-BURP",
      "pattern": "(?i)burpcollaborator\\.net",
      "title": "Burp Collaborator (pentest C2)",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • interact.sh (OOB exfil)rule:c2
    RECIPE-C2-INTERACTSH
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-INTERACTSH",
      "pattern": "(?i)interact\\.sh",
      "title": "interact.sh (OOB exfil)",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • oast.fun (OOB testing)rule:c2
    RECIPE-C2-OAST
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-OAST",
      "pattern": "(?i)(?:^|[^a-zA-Z0-9-])oast\\.fun(?:[^a-zA-Z0-9.-]|$)",
      "title": "oast.fun (OOB testing)",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • Canary Tokensrule:c2
    RECIPE-C2-CANARY
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity MEDIUM.

    YAML
    {
      "id": "C2-CANARY",
      "pattern": "(?i)canarytokens\\.com",
      "title": "Canary Tokens",
      "severity": "MEDIUM",
      "confidence": 0.75,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • Pastebin raw fetchrule:c2
    RECIPE-C2-PASTEBIN
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity MEDIUM.

    YAML
    {
      "id": "C2-PASTEBIN",
      "pattern": "(?i)pastebin\\.com/raw/",
      "title": "Pastebin raw fetch",
      "severity": "MEDIUM",
      "confidence": 0.7,
      "tags": [
        "exfiltration",
        "c2"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • AWS metadata endpoint (SSRF)rule:c2
    RECIPE-C2-METADATA-AWS
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity CRITICAL.

    YAML
    {
      "id": "C2-METADATA-AWS",
      "expression": "f.commands.exists(c, c.argv_complete && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_FETCH in c.operations && f.network.exists(n, n.command_id == c.id && n.action in [defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_DOWNLOAD, defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_CONNECT]))",
      "pattern": "169\\.254\\.169\\.254",
      "title": "AWS metadata endpoint (SSRF)",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "ssrf",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • GCP metadata endpoint (SSRF)rule:c2
    RECIPE-C2-METADATA-GCP
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity CRITICAL.

    YAML
    {
      "id": "C2-METADATA-GCP",
      "pattern": "(?i)(?:^|[^a-zA-Z0-9-])metadata\\.google\\.internal(?:[^a-zA-Z0-9.-]|$)",
      "title": "GCP metadata endpoint (SSRF)",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "ssrf",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • Azure metadata endpoint (SSRF)rule:c2
    RECIPE-C2-METADATA-AZURE
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity CRITICAL.

    YAML
    {
      "id": "C2-METADATA-AZURE",
      "pattern": "169\\.254\\.169\\.254/metadata",
      "title": "Azure metadata endpoint (SSRF)",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "ssrf",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • AWS metadata endpoint (hex-encoded SSRF)rule:c2
    RECIPE-C2-METADATA-HEX
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity CRITICAL.

    YAML
    {
      "id": "C2-METADATA-HEX",
      "pattern": "(?i)0xa9fea9fe",
      "title": "AWS metadata endpoint (hex-encoded SSRF)",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "ssrf",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • AWS metadata endpoint (decimal-encoded SSRF)rule:c2
    RECIPE-C2-METADATA-DECIMAL
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity CRITICAL.

    YAML
    {
      "id": "C2-METADATA-DECIMAL",
      "pattern": "(?:^|[/])2852039166(?:$|[/])",
      "title": "AWS metadata endpoint (decimal-encoded SSRF)",
      "severity": "CRITICAL",
      "confidence": 0.93,
      "tags": [
        "ssrf",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • AWS metadata endpoint (octal-encoded SSRF)rule:c2
    RECIPE-C2-METADATA-OCTAL
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity CRITICAL.

    YAML
    {
      "id": "C2-METADATA-OCTAL",
      "pattern": "0251\\.0376\\.0251\\.0376",
      "title": "AWS metadata endpoint (octal-encoded SSRF)",
      "severity": "CRITICAL",
      "confidence": 0.93,
      "tags": [
        "ssrf",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • DNS TXT query with high-entropy label (tunneling indicator)rule:c2
    RECIPE-C2-DNS-TUNNEL
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-DNS-TUNNEL",
      "pattern": "(?i)\\b(?:dig|host|nslookup|drill)\\b\\s+[^;\\n]*\\$\\((?:whoami|hostname|id(?:\\s+-u)?|cat\\s+(?:/etc/hostname|/etc/machine-id))\\)[A-Za-z0-9._-]*\\.[A-Za-z0-9-]+(?:\\.[A-Za-z0-9-]+)+",
      "title": "DNS TXT query with high-entropy label (tunneling indicator)",
      "severity": "HIGH",
      "confidence": 0.78,
      "tags": [
        "exfiltration",
        "dns-tunnel"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • nslookup with hex subdomain (DNS exfil)rule:c2
    RECIPE-C2-DNS-EXFIL
    egress_external

    Pattern shipped in the bundled strict rule pack (c2.yaml). Severity HIGH.

    YAML
    {
      "id": "C2-DNS-EXFIL",
      "pattern": "(?i)\\bnslookup\\b\\s+[a-f0-9]{8,}\\.\\w+\\.",
      "title": "nslookup with hex subdomain (DNS exfil)",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "exfiltration",
        "dns-tunnel"
      ]
    }
    policies/guardrail/strict/rules/c2.yaml
  • SOUL.md access (agent identity)rule:cognitive-file
    RECIPE-COG-SOUL
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity CRITICAL.

    YAML
    {
      "id": "COG-SOUL",
      "pattern": "(?i)SOUL\\.md",
      "title": "SOUL.md access (agent identity)",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • IDENTITY.md accessrule:cognitive-file
    RECIPE-COG-IDENTITY
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity CRITICAL.

    YAML
    {
      "id": "COG-IDENTITY",
      "pattern": "(?i)IDENTITY\\.md",
      "title": "IDENTITY.md access",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • MEMORY.md mutation or referencerule:cognitive-file
    RECIPE-COG-MEMORY
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "COG-MEMORY",
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)MEMORY\\.md",
      "title": "MEMORY.md mutation or reference",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • CLAUDE.md accessrule:cognitive-file
    RECIPE-COG-CLAUDE-MD
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "COG-CLAUDE-MD",
      "pattern": "(?i)CLAUDE\\.md",
      "title": "CLAUDE.md access",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • TOOLS.md accessrule:cognitive-file
    RECIPE-COG-TOOLS-MD
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "COG-TOOLS-MD",
      "pattern": "(?i)TOOLS\\.md",
      "title": "TOOLS.md access",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • AGENTS.md mutation or referencerule:cognitive-file
    RECIPE-COG-AGENTS-MD
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "COG-AGENTS-MD",
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)AGENTS\\.md",
      "title": "AGENTS.md mutation or reference",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • openclaw.json config accessrule:cognitive-file
    RECIPE-COG-OPENCLAW-JSON
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "COG-OPENCLAW-JSON",
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)openclaw\\.json",
      "title": "openclaw.json config access",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • gateway.json config accessrule:cognitive-file
    RECIPE-COG-GATEWAY-JSON
    sensitive_access

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "COG-GATEWAY-JSON",
      "pattern": "(?i)gateway\\.json",
      "title": "gateway.json config access",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "cognitive-tampering"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • DefenseClaw security state mutationrule:cognitive-file
    RECIPE-tamper.detector_state_write

    Pattern shipped in the bundled strict rule pack (cognitive.yaml). Severity HIGH.

    YAML
    {
      "id": "tamper.detector_state_write",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)(?:~|\\$HOME|\\$\\{HOME\\}|/root|/(?:home|Users)/[^/\\\\\"']+|[a-z]:[/\\\\]Users[/\\\\][^/\\\\\"']+)[/\\\\]\\.defenseclaw[/\\\\](?:\\.env|config\\.yaml|confidence_policy\\.yaml|guardrail_runtime\\.json|device\\.key|picked_connector|custom-providers\\.json|(?:audit|judge|judge_bodies)\\.db(?:-(?:wal|shm))?|(?:cache|policies|quarantine|receipts|registries)[/\\\\][^;\\s]+)\\b",
      "title": "DefenseClaw security state mutation",
      "severity": "HIGH",
      "confidence": 0.98,
      "tags": [
        "cognitive-tampering",
        "tampering",
        "security-state"
      ]
    }
    policies/guardrail/strict/rules/cognitive.yaml
  • Bash reverse shellrule:command
    RECIPE-CMD-REVSHELL-BASH

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-REVSHELL-BASH",
      "expression": "f.commands.exists(c, c.program in ['nc', 'nc.exe', 'ncat', 'ncat.exe', 'netcat', 'socat'] && (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONNECT in c.operations || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_LISTEN in c.operations) && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_EXECUTE) && f.network.exists(n, n.command_id == c.id && n.action in [defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_CONNECT, defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_LISTEN] && n.scope in [defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_PUBLIC, defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_UNKNOWN]))",
      "pattern": "(?i)bash\\s+-i\\s+>&\\s*/dev/tcp/",
      "title": "Bash reverse shell",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "execution",
        "reverse-shell"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Reverse shell via /dev/tcprule:command
    RECIPE-CMD-REVSHELL-DEVTCP

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-REVSHELL-DEVTCP",
      "pattern": "/dev/tcp/(?:\\d{1,3}\\.\\d{1,3}|[A-Za-z0-9-]+(?:\\.[A-Za-z0-9-]+)+)/\\d+\\b",
      "title": "Reverse shell via /dev/tcp",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "execution",
        "reverse-shell"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Netcat reverse shell with -erule:command
    RECIPE-CMD-REVSHELL-NC

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-REVSHELL-NC",
      "pattern": "(?i)\\b(?:nc|ncat|netcat)\\b\\s+(?:(?:-[a-zA-Z]*\\s+)*\\S+\\s+\\d+\\s*(?:-e|--exec)\\b|(?:-[a-zA-Z]*\\s+)*(?:-e|--exec)\\s+\\S+\\s+\\S+\\s+\\d+\\b)",
      "title": "Netcat reverse shell with -e",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "execution",
        "reverse-shell"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Python reverse shellrule:command
    RECIPE-CMD-REVSHELL-PYTHON

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-REVSHELL-PYTHON",
      "pattern": "(?is)\\bpython[23]?\\s+-c\\s+.{0,800}?\\bsocket\\s*\\.\\s*socket\\s*\\(\\s*(?:(?:socket\\s*\\.\\s*)?AF_INET6?(?:\\s*,[^)]{0,160})?)?\\s*\\).{0,800}?\\.\\s*connect\\s*\\(\\s*\\(\\s*[\"'](?:(?:\\d{1,3}\\.){3}\\d{1,3}|[A-Za-z0-9-]+(?:\\.[A-Za-z0-9-]+)+)[\"']\\s*,\\s*\\d{1,5}\\s*\\)\\s*\\).{0,800}?(?:os\\s*\\.\\s*dup2\\s*\\(|(?:sys\\s*\\.\\s*)?(?:stdin|stdout|stderr)\\s*=.{0,120}?(?:makefile|fdopen|reopen)).{0,800}?(?:pty\\s*\\.\\s*spawn|subprocess\\s*\\.\\s*(?:Popen|call|run)|os\\s*\\.\\s*(?:system|exec[a-z]*))\\s*\\(",
      "title": "Python reverse shell",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "execution",
        "reverse-shell"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • curl piped to shellrule:command
    RECIPE-CMD-PIPE-CURL
    egress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-PIPE-CURL",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['curl', 'curl.exe', 'invoke-webrequest', 'iwr', 'invoke-restmethod', 'irm'] && (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_FETCH in c.operations || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in c.operations))",
      "pattern": "(?i)(?:\\bcurl\\b\\s+[^|;\\r\\n]*\\|\\s*(?:[/\\w.-]+/)?(?:bash|zsh|sh|dash|ksh|python(?:[0-9.]*)?|perl|ruby)\\b|\\b(?:invoke-webrequest|iwr|invoke-restmethod|irm)\\b[^|;\\r\\n]*\\|\\s*(?:invoke-expression|iex)\\b)",
      "title": "curl piped to shell",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "execution",
        "download-exec"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • wget piped to shellrule:command
    RECIPE-CMD-PIPE-WGET
    egress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-PIPE-WGET",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['wget', 'wget.exe'] && (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_FETCH in c.operations || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in c.operations))",
      "pattern": "(?i)\\bwget\\b\\s+[^|]*\\|\\s*(?:[/\\w.-]+/)?(?:bash|zsh|sh|dash|ksh|python(?:[0-9.]*)?|perl|ruby)\\b",
      "title": "wget piped to shell",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "execution",
        "download-exec"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • base64 decode piped to shellrule:command
    RECIPE-CMD-PIPE-BASE64

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-PIPE-BASE64",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['base64', 'base64.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DECODE in c.operations)",
      "pattern": "(?i:\\bbase64\\b)[^|;\\n]*\\s(?:-d+i?|--decode)\\b[^|;\\n]*\\|\\s*(?i:(?:[/\\w.-]+/)?(?:bash|zsh|sh|dash|ksh|python(?:[0-9.]*)?|perl|ruby))\\b",
      "title": "base64 decode piped to shell",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "execution",
        "obfuscation"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Remote payload execution with anti-forensic cleanuprule:command
    RECIPE-CMD-REMOTE-PAYLOAD-EXEC-CLEANUP

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-REMOTE-PAYLOAD-EXEC-CLEANUP",
      "pattern": "(?is)\\b(?:curl|wget)\\b[^;\\r\\n]*https?://(?:\\d{1,3}\\.){3}\\d{1,3}(?::\\d+)?/[^\\s;]+(?:;|\\s{2,})[^\\r\\n]*?\\bchmod\\s+(?:\\+x|[0-7]*7[0-7]*)\\s+\\S+(?:;|\\s{2,})[^\\r\\n]*?(?:\\./|(?:ba)?sh\\s+)\\S+(?:[ \\t][^;\\r\\n \\t]+)*(?:;|[ \\t]{2,})[^\\r\\n]*?(?:rm\\s+-[a-zA-Z]*r[a-zA-Z]*f[a-zA-Z]*\\s+(?:\\*|[A-Za-z0-9_.-]+\\*)|history\\s+-c)",
      "title": "Remote payload execution with anti-forensic cleanup",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "execution",
        "download-exec",
        "destructive"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Literal public-IP download executed from the same pathrule:command
    RECIPE-exec.remote_ip_download_execute_same_artifact

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "exec.remote_ip_download_execute_same_artifact",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['curl', 'wget'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_FETCH in c.operations)",
      "pattern": "(?is)\\b(?:curl|wget)\\b[^\\r\\n]{0,900}?https?://(?:\\d{1,3}\\.){3}\\d{1,3}(?::\\d+)?/[^\\s;|&]+[\\s\\S]{0,700}?(?:(?:bash|sh)\\s+)?/(?:tmp|var/tmp|dev/shm)/[^\\s;|&]+",
      "title": "Literal public-IP download executed from the same path",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "execution",
        "download-exec",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Shell eval with dynamic inputrule:command
    RECIPE-CMD-EVAL

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-EVAL",
      "pattern": "(?i)(?:^|[;&|\\r\\n]\\s*)eval\\s+[\"'\\$\\(]",
      "title": "Shell eval with dynamic input",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "execution"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Recursive force delete from critical root pathrule:command
    RECIPE-CMD-RM-RF

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-RM-RF",
      "expression": "f.commands.exists(c, c.argv_complete && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DELETE in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE))",
      "pattern": "(?i)\\brm\\s+(?:-[a-zA-Z]*\\s+)*(?:-[a-zA-Z]*)?(?:r[a-zA-Z]*f|f[a-zA-Z]*r)\\b(?:\\s+\\S+)*\\s+/(?:$|[\"'\\s,}\\]]|(?:etc|bin|sbin|usr|var|home|root|opt|boot|lib(?:64)?|srv|mnt|dev|proc|sys)/?(?:$|[\"'\\s,}\\]]))",
      "title": "Recursive force delete from critical root path",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "destructive"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Filesystem format commandrule:command
    RECIPE-CMD-MKFS

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-MKFS",
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['mkfs', 'mkfs.ext2', 'mkfs.ext3', 'mkfs.ext4', 'mke2fs', 'mkfs.xfs', 'mkfs.btrfs', 'mkfs.f2fs', 'mkfs.vfat', 'mkdosfs', 'mkfs.ntfs', 'mkntfs', 'mkfs.minix', 'mkswap', 'mkfs.exfat', 'mkexfatfs'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DISK_WRITE in c.operations && f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND] && p.flavor == defenseclaw.guardrail.semantic.v1.PathFlavor.PATH_FLAVOR_DEVICE))",
      "pattern": "(?i)(?:^|[;&|\\r\\n]\\s*)(?:sudo\\s+)?(?:mkfs(?:\\.(?:ext2|ext3|ext4|xfs|btrfs|f2fs|vfat|ntfs|exfat|minix))?|mke2fs|mkdosfs|mkntfs|mkswap|mkexfatfs)(?:\\s|$)",
      "title": "Filesystem format command",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "destructive"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Destructive write to a disk devicerule:command
    RECIPE-CMD-DEVICE-WIPE

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-DEVICE-WIPE",
      "expression": "f.commands.exists(c, c.argv_complete && c.program != 'dd' && !(c.program in ['mkfs', 'mkfs.ext2', 'mkfs.ext3', 'mkfs.ext4', 'mke2fs', 'mkfs.xfs', 'mkfs.btrfs', 'mkfs.f2fs', 'mkfs.vfat', 'mkdosfs', 'mkfs.ntfs', 'mkntfs', 'mkfs.minix', 'mkswap', 'mkfs.exfat', 'mkexfatfs']) && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DISK_WRITE in c.operations && f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND] && p.flavor == defenseclaw.guardrail.semantic.v1.PathFlavor.PATH_FLAVOR_DEVICE))",
      "pattern": "a^",
      "title": "Destructive write to a disk device",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "destructive"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • dd disk writerule:command
    RECIPE-CMD-DD-IF

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-DD-IF",
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'dd' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DISK_WRITE in c.operations && f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND] && p.flavor == defenseclaw.guardrail.semantic.v1.PathFlavor.PATH_FLAVOR_DEVICE))",
      "pattern": "(?i)\\bdd\\s+if=",
      "title": "dd disk write",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "destructive"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • chmod world-writablerule:command
    RECIPE-CMD-CHMOD-WORLD

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-CHMOD-WORLD",
      "expression": "f.commands.exists(c, c.argv_complete && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_PERMISSION_CHANGE in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_METADATA))",
      "pattern": "(?i)\\bchmod\\s+[0-7]*[0-7][0-7][2367]\\s",
      "title": "chmod world-writable",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "privilege"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • chown to rootrule:command
    RECIPE-CMD-CHOWN-ROOT

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-CHOWN-ROOT",
      "pattern": "a^",
      "title": "chown to root",
      "severity": "HIGH",
      "confidence": 0.75,
      "tags": [
        "privilege"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • sudo invocationrule:command
    RECIPE-CMD-SUDO

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity LOW.

    YAML
    {
      "id": "CMD-SUDO",
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'sudo' && ('-l' in c.argv || '--list' in c.argv || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_PRIVILEGE in c.operations))",
      "pattern": "(?i)\\bsudo\\s+",
      "title": "sudo invocation",
      "severity": "LOW",
      "confidence": 0.5,
      "tags": [
        "privilege"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Write redirect to /etc/rule:command
    RECIPE-CMD-ETC-WRITE

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-ETC-WRITE",
      "pattern": "(?i)>\\s*/etc/",
      "title": "Write redirect to /etc/",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "system-file"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Crontab modificationrule:command
    RECIPE-CMD-CRONTAB

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-CRONTAB",
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'crontab' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_SCHEDULE in c.operations)",
      "pattern": "(?i)(?:^|[;&|\\r\\n]\\s*)(?:sudo\\s+)?crontab(?:\\s+-u\\s+\\S+)?\\s+(?:-e|-r|-)(?:\\s|$)",
      "title": "Crontab modification",
      "severity": "HIGH",
      "confidence": 0.75,
      "tags": [
        "persistence"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Windows registry persistence modificationrule:command
    RECIPE-CMD-WIN-REG-PERSIST

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-WIN-REG-PERSIST",
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['reg', 'reg.exe', 'set-itemproperty', 'sp', 'new-itemproperty'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations)",
      "pattern": "(?i)\\b(?:reg(?:\\.exe)?\\s+add|(?:set|new)-itemproperty)\\b",
      "title": "Windows registry persistence modification",
      "severity": "CRITICAL",
      "confidence": 0.97,
      "tags": [
        "persistence",
        "windows"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Bounded SAM and SYSTEM registry hive dumprule:command
    RECIPE-credential.windows_registry_hive_dump

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "credential.windows_registry_hive_dump",
      "tool_call_only": true,
      "expression": "f.paths.exists(p, p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ && p.flavor == defenseclaw.guardrail.semantic.v1.PathFlavor.PATH_FLAVOR_REGISTRY && p.normalized in ['HKLM/SAM', 'HKLM/sam']) && f.paths.exists(p, p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ && p.flavor == defenseclaw.guardrail.semantic.v1.PathFlavor.PATH_FLAVOR_REGISTRY && p.normalized in ['HKLM/SYSTEM', 'HKLM/system'])",
      "pattern": "(?i)(?:\\breg(?:\\.exe)?\\s+(?:save|export)\\s+(?:HKLM|HKEY_LOCAL_MACHINE)[\\\\/]SAM\\b[\\s\\S]{0,600}\\breg(?:\\.exe)?\\s+(?:save|export)\\s+(?:HKLM|HKEY_LOCAL_MACHINE)[\\\\/]SYSTEM\\b|\\breg(?:\\.exe)?\\s+(?:save|export)\\s+(?:HKLM|HKEY_LOCAL_MACHINE)[\\\\/]SYSTEM\\b[\\s\\S]{0,600}\\breg(?:\\.exe)?\\s+(?:save|export)\\s+(?:HKLM|HKEY_LOCAL_MACHINE)[\\\\/]SAM\\b)",
      "title": "Bounded SAM and SYSTEM registry hive dump",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "credential",
        "windows",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Dump Windows LSASS credential memoryrule:command
    RECIPE-credential.windows_lsass_memory_dump

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "credential.windows_lsass_memory_dump",
      "tool_call_only": true,
      "pattern": "(?is)(?:\\bprocdump(?:\\.exe)?\\b.{0,240}?(?:-ma|-mm)\\s+lsass(?:\\.exe)?\\s+\\S+\\.dmp\\b|\\bget-process\\s+lsass(?:\\.exe)?\\s*\\|\\s*out-minidump\\b|\\brundll32(?:\\.exe)?\\b[^\\r\\n]{0,260}\\bcomsvcs\\.dll\\s*,?\\s+minidump\\s+\\(\\s*get-process\\s+lsass(?:\\.exe)?\\s*\\)\\.id\\s+\\S{1,260}\\s+full\\b)",
      "title": "Dump Windows LSASS credential memory",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "credential",
        "windows",
        "credential-dumping"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Delete all Windows Volume Shadow Copiesrule:command
    RECIPE-impact.windows_delete_all_shadow_copies

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.windows_delete_all_shadow_copies",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['vssadmin', 'vssadmin.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DELETE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations)",
      "pattern": "(?i)\\bvssadmin(?:\\.exe)?\\s+delete\\s+shadows\\s+(?:/all\\s+/quiet|/quiet\\s+/all)(?:\\s|$)",
      "title": "Delete all Windows Volume Shadow Copies",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "impact",
        "destructive",
        "windows",
        "recovery-inhibition"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Delete a Windows NTFS USN change journalrule:command
    RECIPE-tamper.windows_usn_journal_delete

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.windows_usn_journal_delete",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['fsutil', 'fsutil.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DELETE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations)",
      "pattern": "(?i)\\bfsutil(?:\\.exe)?\\s+usn\\s+deletejournal\\s+/d\\s+[a-z]:(?:\\s|$)",
      "title": "Delete a Windows NTFS USN change journal",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "logs"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable Windows boot recovery and failure handlingrule:command
    RECIPE-impact.windows_recovery_disable_pair

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.windows_recovery_disable_pair",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['bcdedit', 'bcdedit.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\bbcdedit(?:\\.exe)?\\s+/set\\s+\\{[^}\\s]+\\}\\s+bootstatuspolicy\\s+ignoreallfailures\\b[\\s\\S]{0,500}\\bbcdedit(?:\\.exe)?\\s+/set\\s+\\{[^}\\s]+\\}\\s+recoveryenabled\\s+no\\b|\\bbcdedit(?:\\.exe)?\\s+/set\\s+\\{[^}\\s]+\\}\\s+recoveryenabled\\s+no\\b[\\s\\S]{0,500}\\bbcdedit(?:\\.exe)?\\s+/set\\s+\\{[^}\\s]+\\}\\s+bootstatuspolicy\\s+ignoreallfailures\\b)",
      "title": "Disable Windows boot recovery and failure handling",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "impact",
        "destructive",
        "windows",
        "recovery-inhibition",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Clear Windows audit policy and all-user policyrule:command
    RECIPE-tamper.windows_audit_policy_wipe

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.windows_audit_policy_wipe",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['auditpol', 'auditpol.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\bauditpol(?:\\.exe)?\\s+/clear\\s+/y\\b[\\s\\S]{0,400}\\bauditpol(?:\\.exe)?\\s+/remove\\s+/allusers\\b|\\bauditpol(?:\\.exe)?\\s+/remove\\s+/allusers\\b[\\s\\S]{0,400}\\bauditpol(?:\\.exe)?\\s+/clear\\s+/y\\b)",
      "title": "Clear Windows audit policy and all-user policy",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Destroy Windows recovery storesrule:command
    RECIPE-impact.windows_recovery_store_destruction

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.windows_recovery_store_destruction",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['wbadmin', 'wbadmin.exe', 'wmic', 'wmic.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DELETE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations)",
      "pattern": "(?i)(?:\\bwbadmin(?:\\.exe)?\\s+delete\\s+(?:systemstatebackup\\s+-keepversions:0|catalog\\s+-quiet)(?:\\s|$)|\\bwmic(?:\\.exe)?\\s+shadowcopy\\s+delete(?:\\s|$))",
      "title": "Destroy Windows recovery stores",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "impact",
        "destructive",
        "windows",
        "recovery-inhibition"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable multiple Windows Defender controlsrule:command
    RECIPE-tamper.windows_defender_multi_control_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.windows_defender_multi_control_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'set-mppreference' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)\\bSet-MpPreference\\b[^\\r\\n]*(?:-DisableRealtimeMonitoring|-drtm|-DisableBehaviorMonitoring|-dbm|-DisableScriptScanning|-dscrptsc|-DisableBlockAtFirstSeen|-dbaf)\\s+(?:1|true|\\$true)\\b",
      "title": "Disable multiple Windows Defender controls",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable an exact Windows Defender componentrule:command
    RECIPE-tamper.windows_defender_component_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.windows_defender_component_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['mpcmdrun', 'mpcmdrun.exe', 'dism', 'dism.exe', 'disable-windowsoptionalfeature', 'remove-windowsfeature', 'uninstall-windowsfeature', 'schtasks', 'schtasks.exe', 'add-dnsclientnrptrule'])",
      "pattern": "(?i)(?:\\bmpcmdrun(?:\\.exe)?\\b[^\\r\\n]*-RemoveDefinitions[^\\r\\n]*-All\\b|\\bdism(?:\\.exe)?\\b[^\\r\\n]*/Disable-Feature[^\\r\\n]*/FeatureName:Windows-Defender\\b|\\b(?:Disable-WindowsOptionalFeature|Remove-WindowsFeature|Uninstall-WindowsFeature)\\b[^\\r\\n]*Windows-Defender\\b|\\bschtasks(?:\\.exe)?\\b[^\\r\\n]*/delete[^\\r\\n]*Windows Defender[^\\r\\n]*/f\\b|\\bAdd-DnsClientNrptRule\\b[^\\r\\n]*endpoint\\.security\\.microsoft\\.com[^\\r\\n]*127\\.0\\.0\\.1\\b)",
      "title": "Disable an exact Windows Defender component",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "endpoint-security"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable Windows security controls through the registryrule:command
    RECIPE-tamper.windows_registry_security_control_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.windows_registry_security_control_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['reg', 'reg.exe', 'set-itemproperty', 'new-itemproperty'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations)",
      "pattern": "(?i)(?:\\breg(?:\\.exe)?\\s+add\\b[^\\r\\n]*(?:windows defender|windowsfirewall|firewallpolicy|enablelua|consentpromptbehavioradmin|promptonsecuredesktop|hypervisorenforcedcodeintegrity|allowencryptionoracle|rdp-tcp)|\\b(?:set|new)-itemproperty\\b[^\\r\\n]*(?:windows defender|windowsfirewall|enablelua|consentpromptbehavioradmin|promptonsecuredesktop))",
      "title": "Disable Windows security controls through the registry",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "registry",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Hijack a Windows UAC auto-elevation handlerrule:command
    RECIPE-privilege.windows_uac_autoelevation_hijack

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "privilege.windows_uac_autoelevation_hijack",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['reg', 'reg.exe', 'set-itemproperty', 'new-itemproperty', 'new-item']) && f.commands.exists(c, c.argv_complete && c.program == 'start-process')",
      "pattern": "(?is)(?:\\b(?:reg(?:\\.exe)?\\s+add|(?:set|new)-itemproperty)\\b[^\\r\\n]{0,900}(?:mscfile|ms-settings|folder)[/\\\\]shell[/\\\\]open[/\\\\]command[\\s\\S]{0,900}\\b(?:eventvwr(?:\\.msc)?|fodhelper|sdclt)(?:\\.exe)?\\b)",
      "title": "Hijack a Windows UAC auto-elevation handler",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "privilege",
        "defense-evasion",
        "windows",
        "registry",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Hijack a Windows accessibility feature for SYSTEM command accessrule:command
    RECIPE-persistence.windows_accessibility_feature_hijack

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "persistence.windows_accessibility_feature_hijack",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && ((c.program in ['reg', 'reg.exe', 'set-itemproperty', 'new-itemproperty'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations) || (c.program in ['copy', 'copy.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_COPY in c.operations)))",
      "pattern": "(?i)(?:\\b(?:reg(?:\\.exe)?\\s+add|(?:set|new)-itemproperty)\\b[^\\r\\n]*image file execution options[/\\\\](?:osk|sethc|utilman|magnify|narrator|displayswitch|atbroker)\\.exe[^\\r\\n]*\\bDebugger\\b[^\\r\\n]*windows[/\\\\]system32[/\\\\]cmd\\.exe\\b|\\bcopy(?:\\.exe)?\\s+(?:/Y\\s+)?[A-Z]:\\\\Windows\\\\System32\\\\cmd\\.exe\\s+[A-Z]:\\\\Windows\\\\System32\\\\(?:osk|sethc|utilman|magnify|narrator|displayswitch|atbroker)\\.exe\\b)",
      "title": "Hijack a Windows accessibility feature for SYSTEM command access",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "persistence",
        "privilege",
        "windows",
        "registry"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable exact Windows ETW or event telemetry controlsrule:command
    RECIPE-tamper.windows_telemetry_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "tamper.windows_telemetry_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['reg', 'reg.exe', 'set-itemproperty', 'new-itemproperty'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\breg(?:\\.exe)?\\s+add\\b[^\\r\\n]*(?:COMPlus_ETWEnabled|ETWEnabled|Microsoft-Windows-Windows Defender[/\\\\]Operational|Autologger[/\\\\]EventLog-Application)|\\b(?:set|new)-itemproperty\\b[^\\r\\n]*(?:COMPlus_ETWEnabled|ETWEnabled|Microsoft-Windows-Windows Defender[/\\\\]Operational|Autologger[/\\\\]EventLog-Application))",
      "title": "Disable exact Windows ETW or event telemetry controls",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "telemetry",
        "registry"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Weaken exact Windows credential-protection controlsrule:command
    RECIPE-tamper.windows_credential_protection_weaken

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "tamper.windows_credential_protection_weaken",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['reg', 'reg.exe', 'set-itemproperty', 'new-itemproperty'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\breg(?:\\.exe)?\\s+add\\b[^\\r\\n]*(?:UseLogonCredential|RunAsPPL|AutoAdminLogon|DefaultPassword)|\\b(?:set|new)-itemproperty\\b[^\\r\\n]*(?:UseLogonCredential|RunAsPPL|AutoAdminLogon|DefaultPassword))",
      "title": "Weaken exact Windows credential-protection controls",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "credential",
        "defense-evasion",
        "windows",
        "registry"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable or bypass exact Windows AMSI controlsrule:command
    RECIPE-tamper.windows_amsi_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "tamper.windows_amsi_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['new-itemproperty', 'set-itemproperty', 'remove-item', 'powershell-amsi-reflection'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\b(?:set|new)-itemproperty\\b[^\\r\\n]*AmsiEnable|\\bremove-item\\b[^\\r\\n]*Microsoft[/\\\\]AMSI[/\\\\]Providers[/\\\\]\\{[0-9a-f-]{36}\\}|AmsiUtils[^\\r\\n]*amsiInitFailed[^\\r\\n]*SetValue\\s*\\(\\s*\\$null\\s*,\\s*\\$true\\s*\\))",
      "title": "Disable or bypass exact Windows AMSI controls",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "windows",
        "amsi"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable or uninstall an exact endpoint-security productrule:command
    RECIPE-tamper.endpoint_security_product_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.endpoint_security_product_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['systemctl', 'launchctl', 'sysmon', 'sysmon.exe', 'mdatp', 'sc', 'sc.exe'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\b(?:sudo\\s+)?systemctl\\s+(?:stop|disable)\\s+falcon-sensor\\.service\\b|\\b(?:sudo\\s+)?launchctl\\s+unload\\s+/Library/LaunchDaemons/(?:com\\.carbonblack\\.(?:daemon|defense\\.daemon)|at\\.obdev\\.littlesnitchd)\\.plist\\b|\\bsysmon(?:\\.exe)?\\s+-u(?:\\s|$)|\\b(?:sudo\\s+)?mdatp\\s+config\\s+real-time-protection\\s+--value\\s+disabled(?:\\s|$)|\\bsc(?:\\.exe)?\\s+(?:stop|config|query)\\s+WinDefend\\b)",
      "title": "Disable or uninstall an exact endpoint-security product",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "endpoint-security",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Completely relax the host iptables firewallrule:command
    RECIPE-tamper.complete_firewall_relaxation

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.complete_firewall_relaxation",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'iptables' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)(?:sudo\\s+)?iptables\\s+(?:-F(?:\\s+INPUT)?|-P\\s+(?:INPUT|FORWARD|OUTPUT)\\s+ACCEPT)(?:\\s|$)",
      "title": "Completely relax the host iptables firewall",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "firewall",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Destroy an exact cloud audit-logging controlrule:command
    RECIPE-tamper.cloud_audit_control_destruction

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity LOW.

    YAML
    {
      "id": "tamper.cloud_audit_control_destruction",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['aws', 'gcloud'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_DELETE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\baws\\s+cloudtrail\\s+(?:stop-logging|delete-trail)\\b|\\bgcloud\\s+logging\\s+logs\\s+delete\\s+projects/[a-z][a-z0-9-]{4,28}[a-z0-9]/logs/cloudaudit\\.googleapis\\.com%2Factivity\\s+--quiet(?:\\s|$))",
      "title": "Destroy an exact cloud audit-logging control",
      "severity": "LOW",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "cloud",
        "audit-logging",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • PostgreSQL COPY PROGRAM server-side command executionrule:command
    RECIPE-exec.postgresql_copy_program

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "exec.postgresql_copy_program",
      "tool_call_only": true,
      "expression": "f.tool == 'sql_query'",
      "pattern": "a^",
      "title": "PostgreSQL COPY PROGRAM server-side command execution",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "execution",
        "database",
        "postgresql",
        "server-side",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Attach AWS AdministratorAccess to an exact IAM principalrule:command
    RECIPE-privilege.cloud_iam_administrator_attachment

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity LOW.

    YAML
    {
      "id": "privilege.cloud_iam_administrator_attachment",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'aws' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_PERMISSION_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_PRIVILEGE in c.operations)",
      "pattern": "(?i)\\battach-(?:user|role)-policy\\s+--(?:user|role)-name\\s+[A-Za-z0-9_+=,.@-]{1,64}\\s+--policy-arn\\s+arn:aws:iam::aws:policy/AdministratorAccess(?:\\s|\"|$)",
      "title": "Attach AWS AdministratorAccess to an exact IAM principal",
      "severity": "LOW",
      "confidence": 0.99,
      "tags": [
        "privilege-escalation",
        "cloud",
        "aws",
        "iam",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Create a command-executing SQL UDFrule:command
    RECIPE-exec.sql_command_udf_create

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "exec.sql_command_udf_create",
      "tool_call_only": true,
      "expression": "f.tool in ['sql_query', 'execute_command']",
      "pattern": "a^",
      "title": "Create a command-executing SQL UDF",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "execution",
        "database",
        "mysql",
        "postgresql",
        "udf",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Make an exact Kubernetes CronJob job template privilegedrule:command
    RECIPE-privilege.kubernetes_cronjob_privileged_patch

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "privilege.kubernetes_cronjob_privileged_patch",
      "tool_call_only": true,
      "expression": "f.tool == 'kubectl'",
      "pattern": "a^",
      "title": "Make an exact Kubernetes CronJob job template privileged",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "privilege",
        "kubernetes",
        "cronjob",
        "cluster",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Install exact PAM password-capture configurationrule:command
    RECIPE-credential.pam_password_capture

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "credential.pam_password_capture",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'echo' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)\\becho\\b[^\\r\\n]{0,900}\\bpam_tty_audit\\.so\\b[^\\r\\n]{0,900}>{1,2}\\s*/etc/pam\\.d/[a-z0-9_.-]+(?:\\s|$)",
      "title": "Install exact PAM password-capture configuration",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "credential",
        "credential-capture",
        "linux",
        "tampering",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable exact POSIX logging or host-hardening controlsrule:command
    RECIPE-tamper.posix_logging_hardening_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.posix_logging_hardening_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['sed', 'service', 'sysrc', 'sysctl', 'ufw'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:\\bsed\\s+(?:-i|--in-place)\\s+[\"']?s/Storage=auto/Storage=none/[\"']?\\s+/etc/systemd/journald\\.conf\\b|\\bservice\\s+(?:syslogd|pf)\\s+(?:stop|disable)\\b|\\bsysrc\\s+syslogd_enable=[\"']?NO[\"']?\\b|\\bsysctl\\s+-w\\s+kernel\\.randomize_va_space=0\\b|\\bufw\\s+logging\\s+off\\b)",
      "title": "Disable exact POSIX logging or host-hardening controls",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "linux",
        "freebsd",
        "logs",
        "hardening",
        "bounded-lineage"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Disable a Linux host security controlrule:command
    RECIPE-tamper.linux_security_control_disable

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.linux_security_control_disable",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['auditctl', 'setenforce', 'systemctl'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)(?:auditctl\\s+-e\\s+0|setenforce\\s+(?:0|permissive)|(?:sudo\\s+)?systemctl\\s+stop\\s+(?:auditd|firewalld|systemd-journald|ufw)(?:\\.service)?)(?:\\s|$)",
      "title": "Disable a Linux host security control",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "linux"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Erase all macOS unified logsrule:command
    RECIPE-tamper.macos_unified_log_erase

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.macos_unified_log_erase",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.program in ['log', 'sudo'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)(?:sudo\\s+)?log\\s+erase\\s+--all(?:\\s|$)",
      "title": "Erase all macOS unified logs",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "macos",
        "logs"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Capture credentials with an exact fake macOS update promptrule:command
    RECIPE-credential.macos_fake_update_prompt

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "credential.macos_fake_update_prompt",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'osascript' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CREDENTIAL_READ in c.operations)",
      "pattern": "(?i)\\bosascript\\b[^\\r\\n]*(?:Software Update requires that you type your password to apply changes\\.|To perform a security update MacOS needs your passphrase\\.)[^\\r\\n]*\\bhidden\\s+answer\\b",
      "title": "Capture credentials with an exact fake macOS update prompt",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "credential",
        "input-capture",
        "phishing",
        "macos",
        "atomic-red-team"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Dump decrypted credentials from the macOS login keychainrule:command
    RECIPE-credential.macos_login_keychain_dump

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity LOW.

    YAML
    {
      "id": "credential.macos_login_keychain_dump",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'security' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CREDENTIAL_READ in c.operations)",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)(?:sudo\\s+)?security\\s+dump-keychain\\s+-d\\s+login\\.keychain(?:-db)?(?:\\s|$)",
      "title": "Dump decrypted credentials from the macOS login keychain",
      "severity": "LOW",
      "confidence": 0.99,
      "tags": [
        "credential",
        "credential-dumping",
        "macos",
        "atomic-red-team"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Trigger destructive Linux Magic SysRq actionrule:command
    RECIPE-impact.linux_magic_sysrq_destruction

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.linux_magic_sysrq_destruction",
      "tool_call_only": true,
      "expression": "f.paths.exists(p, p.normalized == '/proc/sysrq-trigger')",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)echo\\s+[\"']?[bocei][\"']?\\s*>\\s*/proc/sysrq-trigger(?:\\s|$)",
      "title": "Trigger destructive Linux Magic SysRq action",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "impact",
        "destructive",
        "linux",
        "host-control"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Power off or reboot the POSIX host with haltrule:command
    RECIPE-impact.posix_host_halt

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.posix_host_halt",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'halt' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations)",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)halt\\s+-(?:p|r)(?:\\s|$)",
      "title": "Power off or reboot the POSIX host with halt",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "impact",
        "destructive",
        "posix",
        "host-control"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Destroy canonical POSIX system logsrule:command
    RECIPE-tamper.posix_system_log_destruction

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "tamper.posix_system_log_destruction",
      "tool_call_only": true,
      "expression": "f.paths.exists(p, p.normalized in ['/var/log/messages', '/var/log/security', '/var/log/syslog', '/var/log/auth.log', '/var/log/secure', '/var/log/audit/audit.log'] && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE])",
      "pattern": "(?i)(?:^|[;\\r\\n]\\s*)(?:cat\\s+/dev/(?:null|zero)\\s*>\\s*/var/log/(?:messages|security|syslog|auth\\.log|secure|audit/audit\\.log)|echo\\s+(?:''|\"\")\\s*>\\s*/var/log/(?:messages|security|syslog|auth\\.log|secure|audit/audit\\.log)|truncate\\s+(?:-s\\s+0|--size=0)\\s+/var/log/(?:messages|security|syslog|auth\\.log|secure|audit/audit\\.log)|(?:rm\\s+-(?:rf|fr)|unlink)\\s+/var/log/(?:messages|security|syslog|auth\\.log|secure|audit/audit\\.log)|(?:sudo\\s+)?journalctl\\s+--vacuum-time=0)(?:\\s|$)",
      "title": "Destroy canonical POSIX system logs",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "defense-evasion",
        "tampering",
        "posix",
        "logs"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Suspicious systemd persistence enablementrule:command
    RECIPE-CMD-SYSTEMCTL

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-SYSTEMCTL",
      "pattern": "(?i)\\bsystemctl\\s+enable\\b(?:\\s+--now\\b)?\\s+\\S*(?:backdoor|payload|persist|reverse|shell|evil)\\S*(?:\\.service)?\\b",
      "title": "Suspicious systemd persistence enablement",
      "severity": "CRITICAL",
      "confidence": 0.82,
      "tags": [
        "persistence"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Netcat listenerrule:command
    RECIPE-CMD-NETCAT-LISTEN

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-NETCAT-LISTEN",
      "pattern": "(?i)(?:^|[;&|\\r\\n]\\s*)(?:nc|ncat|netcat)\\s+(?:\\S+\\s+)*(?:-[a-zA-Z]*l[a-zA-Z]*|--listen)(?:\\s|$)",
      "title": "Netcat listener",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "network",
        "reverse-shell"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • curl file uploadrule:command
    RECIPE-CMD-CURL-UPLOAD
    egress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-CURL-UPLOAD",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ) && f.network.exists(n, n.command_id == c.id && n.action == defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_UPLOAD && n.scope in [defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_PUBLIC, defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_UNKNOWN]) && f.data_flows.exists(d, d.to_command_id == c.id && d.from == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_FILE && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_PROCESS) && f.data_flows.exists(d, d.from_command_id == c.id && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_NETWORK))",
      "pattern": "(?i)\\bcurl\\b\\s+.*(?:--upload-file|-T\\s|--data\\s+@|-F\\s+.*=@)",
      "title": "curl file upload",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "network",
        "exfiltration"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • wget POST data exfilrule:command
    RECIPE-CMD-WGET-POST
    egress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-WGET-POST",
      "pattern": "(?i)\\bwget\\b[^;\\n]*--post-(?:data|file)",
      "title": "wget POST data exfil",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "network",
        "exfiltration"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • socat with EXEC (reverse shell)rule:command
    RECIPE-CMD-SOCAT-EXEC

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "CMD-SOCAT-EXEC",
      "pattern": "(?i)\\bsocat\\b[^;\\n]*\\bEXEC\\b",
      "title": "socat with EXEC (reverse shell)",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "execution",
        "reverse-shell"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Environment variable dumprule:command
    RECIPE-CMD-ENV-DUMP
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "CMD-ENV-DUMP",
      "expression": "f.commands.exists(src, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_ENVIRONMENT_READ in src.operations && f.commands.exists(dst, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in dst.operations && src.pipeline_id != 0 && src.pipeline_id == dst.pipeline_id)) && f.commands.exists(src, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_ENVIRONMENT_READ in src.operations && f.data_flows.exists(d, d.from_command_id == src.id && d.from == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDOUT && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDIN)) && f.commands.exists(dst, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in dst.operations && f.data_flows.exists(d, d.to_command_id == dst.id && d.from == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDOUT && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDIN) && f.network.exists(n, n.command_id == dst.id && n.action == defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_UPLOAD && n.scope in [defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_PUBLIC, defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_UNKNOWN]) && f.data_flows.exists(d, d.from_command_id == dst.id && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_NETWORK))",
      "pattern": "(?i)\\b(?:printenv|export\\s+-p|env)\\b[^|;\\n]*\\|\\s*(?:curl\\b[^;\\n]*(?:--data(?:-binary|-raw|-urlencode)?\\s+@-|-d\\s+@-|(?:--form|-F)\\s+[^;\\n]*=@-|--upload-file\\s+-|-T\\s+-)|wget\\b[^;\\n]*--post-(?:data|file)(?:=|\\s+)@?-)",
      "title": "Environment variable dump",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Cloud secret manager value readrule:command
    RECIPE-secrets.cloud_secret_manager_read
    sensitive_access

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "secrets.cloud_secret_manager_read",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CREDENTIAL_READ in c.operations && c.program in ['aws', 'gcloud', 'az', 'vault', 'op', 'pass', 'security'])",
      "pattern": "(?i)\\b(?:aws\\s+secretsmanager\\s+get-secret-value|gcloud\\s+secrets\\s+versions\\s+access|az\\s+keyvault\\s+secret\\s+show|vault\\s+kv\\s+get|op\\s+(?:read|item\\s+get)|pass\\s+show|security\\s+find-(?:generic|internet)-password)\\b",
      "title": "Cloud secret manager value read",
      "severity": "CRITICAL",
      "confidence": 0.94,
      "tags": [
        "credential",
        "cloud",
        "secret-manager"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Sensitive read piped to external uploadrule:command
    RECIPE-exfil.secret_read_and_egress_oneliner
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "exfil.secret_read_and_egress_oneliner",
      "tool_call_only": true,
      "expression": "f.commands.exists(src, (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CREDENTIAL_READ in src.operations || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in src.operations) && f.paths.exists(p, p.command_id == src.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ)) && f.commands.exists(src, (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CREDENTIAL_READ in src.operations || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in src.operations) && f.commands.exists(dst, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in dst.operations && src.pipeline_id != 0 && src.pipeline_id == dst.pipeline_id)) && f.commands.exists(src, f.data_flows.exists(d, d.from_command_id == src.id && d.from == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDOUT && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDIN)) && f.commands.exists(dst, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_UPLOAD in dst.operations && f.data_flows.exists(d, d.to_command_id == dst.id && d.from == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDOUT && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_STDIN) && f.network.exists(n, n.command_id == dst.id && n.action == defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_UPLOAD && n.scope in [defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_PUBLIC, defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_UNKNOWN]) && f.data_flows.exists(d, d.from_command_id == dst.id && d.to == defenseclaw.guardrail.semantic.v1.DataKind.DATA_KIND_NETWORK))",
      "pattern": "(?i)\\b(?:cat|type|Get-Content)\\b[^|;\\n]*(?:\\.env|credentials|\\.git-credentials|\\.netrc|_netrc|\\.npmrc|\\.pypirc|config\\.json|/token)\\s*\\|\\s*(?:curl\\b[^;\\n]*(?:--data(?:-binary|-raw|-urlencode)?\\s+@-|-d\\s+@-|(?:--form|-F)\\s+[^;\\n]*=@-|--upload-file\\s+-|-T\\s+-)|wget\\b[^;\\n]*--post-(?:data|file)(?:=|\\s+)@?-)",
      "title": "Sensitive read piped to external upload",
      "severity": "CRITICAL",
      "confidence": 0.96,
      "tags": [
        "credential",
        "exfiltration",
        "network"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Unrestricted passwordless sudo grantrule:command
    RECIPE-privilege.sudoers_unrestricted_nopasswd

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "privilege.sudoers_unrestricted_nopasswd",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_WRITE in c.operations || defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_APPEND in c.operations) && f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND]))",
      "pattern": "a^",
      "title": "Unrestricted passwordless sudo grant",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "privilege",
        "persistence",
        "sudoers"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Global dynamic-loader preload installationrule:command
    RECIPE-persistence.global_ld_preload_install

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "persistence.global_ld_preload_install",
      "tool_call_only": true,
      "pattern": "(?i)(?:>|>>|\\btee(?:\\s+-a)?)\\s*/etc/ld\\.so\\.preload\\b",
      "title": "Global dynamic-loader preload installation",
      "severity": "HIGH",
      "confidence": 0.99,
      "tags": [
        "persistence",
        "execution",
        "dynamic-loader"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • SSH reverse tunnelrule:command
    RECIPE-exec.reverse_tunnel
    egress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "exec.reverse_tunnel",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_TUNNEL in c.operations && f.network.exists(n, n.command_id == c.id && n.action in [defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_TUNNEL, defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_CONNECT] && n.scope in [defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_PUBLIC, defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_UNKNOWN]))",
      "pattern": "(?i)(?:\\b(?:ssh|autossh)(?:\\.exe)?\\b[^;\\n]*(?:^|\\s)-R(?:\\s+|\\S)|\\bchisel(?:\\.exe)?\\b[^;\\n]*\\bclient\\b[^;\\n]*\\bR:(?:socks|\\S+)|\\bligolo(?:-ng)?-agent\\b[^;\\n]*--?connect(?:=|\\s+)\\S+)",
      "title": "SSH reverse tunnel",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "network",
        "tunnel",
        "persistence"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Agent runtime safety bypass flagsrule:command
    RECIPE-exec.agent_runtime_bypass_flags

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "exec.agent_runtime_bypass_flags",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['codex', 'codex.exe', 'claude', 'claude.exe', 'gemini', 'gemini.exe', 'opencode', 'opencode.exe', 'npx', 'pnpm', 'bunx'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_POLICY_BYPASS in c.operations)",
      "pattern": "(?i)\\b(?:(?:npx(?:\\s+-y)?|pnpm\\s+dlx|bunx)\\s+)?(?:claude|codex|gemini)(?:\\.exe)?\\b[^;\\n]*(?:--dangerously-skip-permissions|--permission-mode(?:=|\\s+)bypassPermissions\\b|--dangerously-bypass-approvals-and-sandbox|--yolo\\b|(?:--sandbox|-s)(?:=|\\s+)danger-full-access\\b[^;\\n]*(?:--ask-for-approval|-a)(?:=|\\s+)never\\b|(?:--ask-for-approval|-a)(?:=|\\s+)never\\b[^;\\n]*(?:--sandbox|-s)(?:=|\\s+)danger-full-access\\b)",
      "title": "Agent runtime safety bypass flags",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "execution",
        "policy-bypass",
        "agent-runtime"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Git verification hook bypassrule:command
    RECIPE-integrity.git_hooks_bypass

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity MEDIUM.

    YAML
    {
      "id": "integrity.git_hooks_bypass",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'git')",
      "pattern": "(?i)\\bgit\\b[^;\\n]*(?:(?:\\bcommit\\b|\\bpush\\b)[^;\\n]*--no-verify\\b|-c\\s+core\\.hooksPath=(?:/dev/null|nul:?|[\"']{2})[^;\\n]*\\b(?:commit|push)\\b|\\bconfig\\s+core\\.hooksPath\\s+(?:/dev/null|nul:?|[\"']{2}))",
      "title": "Git verification hook bypass",
      "severity": "MEDIUM",
      "confidence": 0.9,
      "tags": [
        "integrity",
        "policy-bypass"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Shell history clear or disablerule:command
    RECIPE-integrity.history_tamper

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity MEDIUM.

    YAML
    {
      "id": "integrity.history_tamper",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && (c.program == 'history' || (c.program == 'unset' && 'HISTFILE' in c.argv)))",
      "pattern": "(?i)(?:\\bhistory\\b[^;\\n]*\\s-[acnrw]*c[acnrw]*(?:[\\s;&|]|$)|\\bunset\\b[^;\\n]*\\bHISTFILE\\b)",
      "title": "Shell history clear or disable",
      "severity": "MEDIUM",
      "confidence": 0.95,
      "tags": [
        "integrity",
        "defense-evasion"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Multi-host network sweeprule:command
    RECIPE-recon.network_sweep
    egress_external

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity MEDIUM.

    YAML
    {
      "id": "recon.network_sweep",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_NETWORK_SCAN in c.operations && f.network.exists(n, n.command_id == c.id && n.action == defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_SCAN && n.target_kind in [defenseclaw.guardrail.semantic.v1.NetworkTargetKind.NETWORK_TARGET_KIND_MULTI_ADDRESS_CIDR, defenseclaw.guardrail.semantic.v1.NetworkTargetKind.NETWORK_TARGET_KIND_RANGE, defenseclaw.guardrail.semantic.v1.NetworkTargetKind.NETWORK_TARGET_KIND_LIST, defenseclaw.guardrail.semantic.v1.NetworkTargetKind.NETWORK_TARGET_KIND_GENERATED]))",
      "pattern": "(?i)\\b(?:nmap|masscan|zmap|rustscan|naabu|fping)\\b[^;\\n]*(?:\\b(?:-iL|--input-file|--target-file|-list|--list|-iR|-g|--generate)\\b|(?:\\d{1,3}\\.){3}\\d{1,3}/(?:[0-9]|[12][0-9]|3[01])\\b)",
      "title": "Multi-host network sweep",
      "severity": "MEDIUM",
      "confidence": 0.9,
      "tags": [
        "reconnaissance",
        "network"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Privileged container with host-root bindrule:command
    RECIPE-privilege.container_host_escape

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "privilege.container_host_escape",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['docker', 'podman', 'nerdctl'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONTAINER_RUN in c.operations && c.argv.exists(a, a == '--privileged' || a.startsWith('--privileged=')) && f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE] && (p.normalized == '/' || p.resolved == '/')))",
      "pattern": "(?i)\\b(?:docker|podman|nerdctl)\\b[^;\\n]*\\brun\\b[^;\\n]*--privileged\\b[^;\\n]*(?:-v|--volume)(?:=|\\s+)/:[^;\\s]+",
      "title": "Privileged container with host-root bind",
      "severity": "CRITICAL",
      "confidence": 0.96,
      "tags": [
        "privilege",
        "container",
        "host-escape"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Host PID-1 namespace or root entryrule:command
    RECIPE-privilege.host_namespace_entry

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "privilege.host_namespace_entry",
      "tool_call_only": true,
      "pattern": "(?i)\\b(?:nsenter\\b[^;\\n]*(?:(?:-t|--target)(?:=|\\s+)1\\b|/proc/1/ns/(?:mnt|uts|ipc|net|pid|user|cgroup|time)\\b)|chroot\\b[^;\\n]*/proc/1/root\\b)",
      "title": "Host PID-1 namespace or root entry",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "privilege",
        "container",
        "host-escape"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Remote workload command executionrule:command
    RECIPE-lateral.workload_exec

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "lateral.workload_exec",
      "tool_call_only": true,
      "pattern": "(?i)\\b(?:(?:kubectl|oc)\\b[^;\\n]*\\b(?:exec|debug|rsh)\\b|crictl\\b[^;\\n]*\\bexec\\b|ctr\\b[^;\\n]*\\btasks?\\s+exec\\b|(?:docker|podman)\\b[^;\\n]*(?:--host|-H)(?:=|\\s+)(?:tcp|ssh)://[^;\\s]+\\s+exec\\b)",
      "title": "Remote workload command execution",
      "severity": "HIGH",
      "confidence": 0.92,
      "tags": [
        "execution",
        "lateral-movement",
        "workload"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Cryptocurrency miner launchrule:command
    RECIPE-impact.cryptomining_launch

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "impact.cryptomining_launch",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['docker', 'podman', 'nerdctl'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONTAINER_RUN in c.operations)",
      "pattern": "(?i)(?:^|[;&|\\s])(?:[/\\\\\\w.-]+[/\\\\])?(?:xmrig(?:-proxy)?|minerd|cpuminer(?:-multi)?|ethminer|cgminer|bfgminer|t-rex|lolminer|nbminer|teamredminer|phoenixminer|nanominer)(?:\\s|$)",
      "title": "Cryptocurrency miner launch",
      "severity": "HIGH",
      "confidence": 0.95,
      "tags": [
        "impact",
        "resource-hijacking"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Unbounded fork bombrule:command
    RECIPE-impact.fork_bomb

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.fork_bomb",
      "tool_call_only": true,
      "pattern": "(?i)(?:^|\\s)(?::\\s*\\(\\s*\\)\\s*\\{\\s*:\\s*\\|\\s*:\\s*&\\s*\\}\\s*;\\s*:|perl\\s+-e\\s+['\"]?\\s*fork\\s+while\\s+fork\\s*;?\\s*['\"]?|ruby\\s+-e\\s+['\"]?\\s*loop\\s*\\{\\s*fork\\s*\\}\\s*;?\\s*['\"]?)\\s*$",
      "title": "Unbounded fork bomb",
      "severity": "CRITICAL",
      "confidence": 0.99,
      "tags": [
        "impact",
        "denial-of-service"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Mass process terminationrule:command
    RECIPE-impact.mass_process_termination

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "impact.mass_process_termination",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['kill', 'stop-process', 'taskkill'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_PROCESS_KILL in c.operations)",
      "pattern": "(?i)\\b(?:kill\\b[^;\\n]*(?:-9|-KILL|-SIGKILL)\\b[^;\\n]*-1\\b|stop-process\\b[^;\\n]*(?:-name|-processname)(?:=|\\s+)\\*[^;\\n]*-force\\b|taskkill\\b[^;\\n]*/im\\s+\\*[^;\\n]*/f\\b)",
      "title": "Mass process termination",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "impact",
        "denial-of-service"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Git remote routing changerule:command
    RECIPE-source.git_remote_tamper

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity MEDIUM.

    YAML
    {
      "id": "source.git_remote_tamper",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'git' && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations && f.network.exists(n, n.command_id == c.id && n.action == defenseclaw.guardrail.semantic.v1.NetworkAction.NETWORK_ACTION_CONNECT && n.scope in [defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_PUBLIC, defenseclaw.guardrail.semantic.v1.NetworkScope.NETWORK_SCOPE_UNKNOWN]))",
      "pattern": "(?i)\\bgit\\b[^;\\n]*\\bremote\\s+(?:add|set-url)\\b[^;\\n]*(?:https?|ssh|git)://[^;\\s]+",
      "title": "Git remote routing change",
      "severity": "MEDIUM",
      "confidence": 0.9,
      "tags": [
        "source-control",
        "network",
        "tampering"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Git executable configuration change or overwriterule:command
    RECIPE-source.git_config_exec

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity MEDIUM.

    YAML
    {
      "id": "source.git_config_exec",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program == 'git' && (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONFIG_CHANGE in c.operations || (defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_WRITE in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE && (p.normalized == '.git/config' || p.resolved.endsWith('/.git/config'))))))",
      "pattern": "(?i)\\bgit\\b[^;\\r\\n]*(?:\\bconfig\\b[^;\\r\\n]*(?:alias\\.[A-Za-z0-9._-]+|credential\\.helper|core\\.(?:sshCommand|hooksPath))\\b[^;\\r\\n]+|\\b(?:show|log|diff|whatchanged)\\b[^;\\r\\n]*--output(?:=|\\s+)[^;\\r\\n]*\\.git[/\\\\]config(?:$|[\"'\\s]))",
      "title": "Git executable configuration change or overwrite",
      "severity": "MEDIUM",
      "confidence": 0.9,
      "tags": [
        "source-control",
        "execution",
        "tampering"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • SSH authorized keys mutationrule:command
    RECIPE-persistence.ssh_authorized_keys_command

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity CRITICAL.

    YAML
    {
      "id": "persistence.ssh_authorized_keys_command",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)(?:>>?|(?:tee|truncate|rm|unlink|cp|mv|set-content|add-content|out-file|remove-item|copy-item|move-item)\\b[^;\\n]*)[^;\\n]*(?:[/\\\\]\\.ssh[/\\\\](?:authorized_keys2?)|[/\\\\]programdata[/\\\\]ssh[/\\\\]administrators_authorized_keys)\\b",
      "title": "SSH authorized keys mutation",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "persistence",
        "credential",
        "ssh"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • Privileged account or group changerule:command
    RECIPE-persistence.privileged_account_change

    Pattern shipped in the bundled strict rule pack (commands.yaml). Severity HIGH.

    YAML
    {
      "id": "persistence.privileged_account_change",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, c.argv_complete && c.program in ['useradd', 'usermod', 'gpasswd', 'groupmems', 'adduser', 'dseditgroup', 'dscl', 'net', 'net1', 'add-localgroupmember', 'add-adgroupmember'] && defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_ACCOUNT_CHANGE in c.operations)",
      "pattern": "(?i)\\b(?:useradd\\b[^;\\n]*(?:-u|--uid)(?:=|\\s+)0\\b|usermod\\b[^;\\n]*(?:(?:-u|--uid)(?:=|\\s+)0\\b|(?:-G|--groups)(?:=|\\s+)[^;\\n]*(?:root|sudo|wheel|admin|docker|lxd|incus|podman))|(?:gpasswd|groupmems|adduser|dseditgroup|dscl)\\b[^;\\n]*(?:root|sudo|wheel|admin|docker|lxd|incus|podman)\\b|net\\s+localgroup\\s+(?:Administrators|Remote Desktop Users)\\b[^;\\n]*/add\\b|Add-(?:Local|AD)GroupMember\\b[^;\\n]*(?:Administrators|Domain Admins|Enterprise Admins)\\b)",
      "title": "Privileged account or group change",
      "severity": "HIGH",
      "confidence": 0.95,
      "tags": [
        "persistence",
        "privilege",
        "account"
      ]
    }
    policies/guardrail/strict/rules/commands.yaml
  • US Social Security Numberrule:enterprise-data
    RECIPE-ENT-BULK-SSN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity CRITICAL.

    YAML
    {
      "id": "ENT-BULK-SSN",
      "pattern": "\\b(?:00[1-9]|0[1-9][0-9]|[1-5][0-9]{2}|6[0-5][0-9]|66[0-5]|66[7-9]|6[7-9][0-9]|[78][0-9]{2})-(?:0[1-9]|[1-9][0-9])-(?:000[1-9]|00[1-9][0-9]|0[1-9][0-9]{2}|[1-9][0-9]{3})\\b",
      "title": "US Social Security Number",
      "severity": "CRITICAL",
      "confidence": 0.85,
      "tags": [
        "pii",
        "regulated"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • US SSN (no hyphens)rule:enterprise-data
    RECIPE-ENT-BULK-SSN-NOHYPHEN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-BULK-SSN-NOHYPHEN",
      "pattern": "\\b(?:00[1-9]|0[1-9][0-9]|[1-5][0-9]{2}|6[0-5][0-9]|66[0-5]|66[7-9]|6[7-9][0-9]|[78][0-9]{2})(?:0[1-9]|[1-9][0-9])(?:000[1-9]|00[1-9][0-9]|0[1-9][0-9]{2}|[1-9][0-9]{3})\\b",
      "title": "US SSN (no hyphens)",
      "severity": "HIGH",
      "confidence": 0.55,
      "tags": [
        "pii",
        "regulated"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • Visa credit card numberrule:enterprise-data
    RECIPE-ENT-CC-VISA
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity CRITICAL.

    YAML
    {
      "id": "ENT-CC-VISA",
      "pattern": "\\b4\\d{3}[\\s-]?\\d{4}[\\s-]?\\d{4}[\\s-]?\\d{4}\\b",
      "title": "Visa credit card number",
      "severity": "CRITICAL",
      "confidence": 0.8,
      "tags": [
        "pii",
        "pci"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • Mastercard credit card numberrule:enterprise-data
    RECIPE-ENT-CC-MC
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity CRITICAL.

    YAML
    {
      "id": "ENT-CC-MC",
      "pattern": "\\b5[1-5]\\d{2}[\\s-]?\\d{4}[\\s-]?\\d{4}[\\s-]?\\d{4}\\b",
      "title": "Mastercard credit card number",
      "severity": "CRITICAL",
      "confidence": 0.8,
      "tags": [
        "pii",
        "pci"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • American Express card numberrule:enterprise-data
    RECIPE-ENT-CC-AMEX
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity CRITICAL.

    YAML
    {
      "id": "ENT-CC-AMEX",
      "pattern": "\\b3[47]\\d{2}[\\s-]?\\d{6}[\\s-]?\\d{5}\\b",
      "title": "American Express card number",
      "severity": "CRITICAL",
      "confidence": 0.8,
      "tags": [
        "pii",
        "pci"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • Discover card numberrule:enterprise-data
    RECIPE-ENT-CC-DISCOVER
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity CRITICAL.

    YAML
    {
      "id": "ENT-CC-DISCOVER",
      "pattern": "\\b6(?:011|5\\d{2})[\\s-]?\\d{4}[\\s-]?\\d{4}[\\s-]?\\d{4}\\b",
      "title": "Discover card number",
      "severity": "CRITICAL",
      "confidence": 0.8,
      "tags": [
        "pii",
        "pci"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • International Bank Account Number (IBAN)rule:enterprise-data
    RECIPE-ENT-IBAN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-IBAN",
      "pattern": "\\b[A-Z]{2}\\d{2}[\\s]?[\\dA-Z]{4}[\\s]?(?:[\\dA-Z]{4}[\\s]?){1,7}[\\dA-Z]{1,4}\\b",
      "title": "International Bank Account Number (IBAN)",
      "severity": "HIGH",
      "confidence": 0.75,
      "tags": [
        "pii",
        "financial"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • US phone numberrule:enterprise-data
    RECIPE-ENT-US-PHONE
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity MEDIUM.

    YAML
    {
      "id": "ENT-US-PHONE",
      "pattern": "(?:\\+1(?:[\\s.-]?\\d){10}\\b|\\(\\d{3}\\)[\\s.-]?\\d{3}[\\s.-]?\\d{4}\\b|\\b\\d{3}[\\s.-]\\d{3}[\\s.-]\\d{4}\\b|(?i:\\b(?:phone|mobile|telephone|contact|tel|whatsapp)\\b[^\\r\\n\\d]{0,32}(?:\\+?1[\\s.-]?)?(?:\\(?\\d{3}\\)?[\\s.-]?)\\d{3}[\\s.-]?\\d{4}\\b))",
      "title": "US phone number",
      "severity": "MEDIUM",
      "confidence": 0.5,
      "tags": [
        "pii"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • International E.164-style phone numberrule:enterprise-data
    RECIPE-ENT-PHONE-E164
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity MEDIUM.

    YAML
    {
      "id": "ENT-PHONE-E164",
      "pattern": "(?:^|[\\s:(])\\+\\d{1,3}(?:[\\s.-]?\\d){7,14}\\b",
      "title": "International E.164-style phone number",
      "severity": "MEDIUM",
      "confidence": 0.55,
      "tags": [
        "pii"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • Email addressrule:enterprise-data
    RECIPE-ENT-EMAIL-BULK
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity MEDIUM.

    YAML
    {
      "id": "ENT-EMAIL-BULK",
      "pattern": "(?i)\\b[a-z0-9._%+\\-]+@[a-z0-9.\\-]+\\.[a-z]{2,}\\b",
      "title": "Email address",
      "severity": "MEDIUM",
      "confidence": 0.4,
      "tags": [
        "pii"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • US passport number patternrule:enterprise-data
    RECIPE-ENT-PASSPORT-US
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-PASSPORT-US",
      "pattern": "\\b[A-Z]\\d{8}\\b",
      "title": "US passport number pattern",
      "severity": "HIGH",
      "confidence": 0.5,
      "tags": [
        "pii",
        "regulated"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • California drivers license patternrule:enterprise-data
    RECIPE-ENT-DL-CA
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-DL-CA",
      "pattern": "\\b[A-Z]\\d{7}\\b",
      "title": "California drivers license pattern",
      "severity": "HIGH",
      "confidence": 0.4,
      "tags": [
        "pii",
        "regulated"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • Medical record numberrule:enterprise-data
    RECIPE-ENT-MEDICAL-RECORD
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity CRITICAL.

    YAML
    {
      "id": "ENT-MEDICAL-RECORD",
      "pattern": "(?i)\\b(?:mrn|medical record|patient id)\\s*[:#]?\\s*\\d{6,12}\\b",
      "title": "Medical record number",
      "severity": "CRITICAL",
      "confidence": 0.7,
      "tags": [
        "pii",
        "hipaa"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • Date of birth with labelrule:enterprise-data
    RECIPE-ENT-DOB-PATTERN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-DOB-PATTERN",
      "pattern": "(?i)\\b(?:dob|date of birth|birth[\\s-]?date)\\s*[:#]?\\s*\\d{1,2}[/\\-]\\d{1,2}[/\\-]\\d{2,4}\\b",
      "title": "Date of birth with label",
      "severity": "HIGH",
      "confidence": 0.75,
      "tags": [
        "pii",
        "hipaa"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • UK NHS number patternrule:enterprise-data
    RECIPE-ENT-NHS-NUMBER
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-NHS-NUMBER",
      "pattern": "(?i:\\b(?:nhs(?:\\s+(?:number|no\\.?|id))?|national health service(?:\\s+(?:number|no\\.?|id))?)\\b[^\\r\\n\\d]{0,24}\\d{3}[\\s]?\\d{3}[\\s]?\\d{4}\\b|\\b\\d{3}[\\s]?\\d{3}[\\s]?\\d{4}\\b[^\\r\\nA-Za-z]{0,8}(?:nhs(?:\\s+(?:number|no\\.?|id))?|national health service(?:\\s+(?:number|no\\.?|id))?)\\b)",
      "title": "UK NHS number pattern",
      "severity": "HIGH",
      "confidence": 0.4,
      "tags": [
        "pii",
        "regulated"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • CSV/TSV records with multiple PII columnsrule:enterprise-data
    RECIPE-ENT-BULK-CSV-PII
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-BULK-CSV-PII",
      "pattern": "(?im)(?:^|\\n)[^\\r\\n]*(?:first[\\s_]?name|last[\\s_]?name|ssn|social[\\s_]?security|credit[\\s_]?card|card[\\s_]?number|account[\\s_]?number)[^\\r\\n]*[,\\t][^\\r\\n]*(?:first[\\s_]?name|last[\\s_]?name|ssn|social[\\s_]?security|credit[\\s_]?card|card[\\s_]?number|account[\\s_]?number)[^\\r\\n]*\\r?\\n[^\\r\\n]*[0-9][^\\r\\n]*",
      "title": "CSV/TSV records with multiple PII columns",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "pii",
        "bulk-data"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • JSON record with multiple PII valuesrule:enterprise-data
    RECIPE-ENT-BULK-JSON-PII
    sensitive_access

    Pattern shipped in the bundled strict rule pack (enterprise-data.yaml). Severity HIGH.

    YAML
    {
      "id": "ENT-BULK-JSON-PII",
      "pattern": "(?i)\\{[^{}]{0,512}\"(?:ssn|social_security|credit_card|card_number|account_number|routing_number)\"\\s*:\\s*(?:\"(?:\\d[\\s-]?){6,19}\"|\\d{6,19}\\b)\\s*,[^{}]{0,512}\"(?:ssn|social_security|credit_card|card_number|account_number|routing_number)\"\\s*:\\s*(?:\"(?:\\d[\\s-]?){6,19}\"|\\d{6,19}\\b)[^{}]{0,512}\\}",
      "title": "JSON record with multiple PII values",
      "severity": "HIGH",
      "confidence": 0.75,
      "tags": [
        "pii",
        "bulk-data"
      ]
    }
    policies/guardrail/strict/rules/enterprise-data.yaml
  • AWS access keyrule:secret
    RECIPE-SEC-AWS-KEY
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    examples (2)
    • AKIAIOSFODNN7EXAMPLE
    • ASIA1234567890ABCDEFGHIJ
    • BANANAFRUITNOTAKEY
    • AKI
    • AKIAtoolow
    YAML
    {
      "id": "SEC-AWS-KEY",
      "pattern": "\\b(?:AKIA|AGPA|AIDA|AROA|AIPA|ANPA|ANVA|ASIA)[0-9A-Z]{16,}",
      "title": "AWS access key",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • AWS secret access keyrule:secret
    RECIPE-SEC-AWS-SECRET
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-AWS-SECRET",
      "pattern": "(?i)aws_secret_access_key\\s*[=:]\\s*[A-Za-z0-9/+=]{30,}",
      "title": "AWS secret access key",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Anthropic API keyrule:secret
    RECIPE-SEC-ANTHROPIC
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-ANTHROPIC",
      "pattern": "sk-ant-[a-zA-Z0-9\\-_]{20,}",
      "title": "Anthropic API key",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • OpenAI project keyrule:secret
    RECIPE-SEC-OPENAI
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-OPENAI",
      "pattern": "sk-proj-[a-zA-Z0-9]{20,}",
      "title": "OpenAI project key",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • OpenAI API key (long form)rule:secret
    RECIPE-SEC-OPENAI-V2
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    examples (1)
    • sk-aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
    • sk-tooshort
    • pk-livenotopenai
    YAML
    {
      "id": "SEC-OPENAI-V2",
      "pattern": "sk-[a-zA-Z0-9]{40,}",
      "title": "OpenAI API key (long form)",
      "severity": "CRITICAL",
      "confidence": 0.85,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Stripe secret/restricted keyrule:secret
    RECIPE-SEC-STRIPE
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-STRIPE",
      "pattern": "(?:sk_live_|sk_test_|rk_live_|rk_test_)[a-zA-Z0-9]{20,}",
      "title": "Stripe secret/restricted key",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • GitHub tokenrule:secret
    RECIPE-SEC-GITHUB-TOKEN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    examples (2)
    • ghp_aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
    • ghs_bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb
    • ghp_short
    • gh_notatoken
    YAML
    {
      "id": "SEC-GITHUB-TOKEN",
      "pattern": "(?:ghp_|gho_|ghu_|ghs_|ghr_)[a-zA-Z0-9]{36,}",
      "title": "GitHub token",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • GitHub fine-grained PATrule:secret
    RECIPE-SEC-GITHUB-PAT
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-GITHUB-PAT",
      "pattern": "github_pat_[a-zA-Z0-9_]{22,}",
      "title": "GitHub fine-grained PAT",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • GitLab personal access tokenrule:secret
    RECIPE-SEC-GITLAB
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-GITLAB",
      "pattern": "glpat-[a-zA-Z0-9\\-_]{20,}",
      "title": "GitLab personal access token",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Google API keyrule:secret
    RECIPE-SEC-GOOGLE
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-GOOGLE",
      "pattern": "AIza[0-9A-Za-z\\-_]{35}",
      "title": "Google API key",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Slack tokenrule:secret
    RECIPE-SEC-SLACK-TOKEN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-SLACK-TOKEN",
      "pattern": "xox[bpors]-[0-9a-zA-Z\\-]{10,}",
      "title": "Slack token",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Slack webhook URLrule:secret
    RECIPE-SEC-SLACK-WEBHOOK
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    examples (1)
    • https://hooks.slack.com/services/T0000/B0000/abcdefg12345
    • https://hooks.slack.com/wrongpath
    • https://example.com
    YAML
    {
      "id": "SEC-SLACK-WEBHOOK",
      "pattern": "https://hooks\\.slack\\.com/services/T[A-Z0-9]+/B[A-Z0-9]+/[a-zA-Z0-9]+",
      "title": "Slack webhook URL",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Discord webhook URLrule:secret
    RECIPE-SEC-DISCORD-WEBHOOK
    sensitive_accessegress_external

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-DISCORD-WEBHOOK",
      "pattern": "https://discord(?:app)?\\.com/api/webhooks/\\d+/[a-zA-Z0-9_\\-]+",
      "title": "Discord webhook URL",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Private keyrule:secret
    RECIPE-SEC-PRIVKEY
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    examples (2)
    • -----BEGIN RSA PRIVATE KEY-----
    • -----BEGIN OPENSSH PRIVATE KEY-----
    • -----BEGIN CERTIFICATE-----
    • BEGIN PRIVATE KEY without dashes
    YAML
    {
      "id": "SEC-PRIVKEY",
      "pattern": "-----BEGIN (?:RSA |EC |OPENSSH |PGP |DSA )?PRIVATE KEY-----",
      "title": "Private key",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • JWT tokenrule:secret
    RECIPE-SEC-JWT
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity MEDIUM.

    examples (1)
    • eyJhbGciOiJIUzI1NiJ9.eyJzdWIiOiIxMjM0NTY3ODkwIn0.signature_here_xyz
    • eyJonly
    • not.a.jwt
    YAML
    {
      "id": "SEC-JWT",
      "pattern": "eyJ[A-Za-z0-9\\-_]{10,}\\.eyJ[A-Za-z0-9\\-_]{10,}\\.[A-Za-z0-9\\-_.+/=]+",
      "title": "JWT token",
      "severity": "MEDIUM",
      "confidence": 0.7,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Connection string with credentialsrule:secret
    RECIPE-SEC-CONNSTR
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-CONNSTR",
      "pattern": "(?:mongodb|postgres|mysql|redis|amqp)://[^:\\s]+:[^@\\s]+@",
      "title": "Connection string with credentials",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Bearer token in headerrule:secret
    RECIPE-SEC-BEARER
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity HIGH.

    YAML
    {
      "id": "SEC-BEARER",
      "pattern": "(?i)(?:authorization|bearer)\\s*[:=]\\s*Bearer\\s+[A-Za-z0-9\\-_.~+/]{16,}=*",
      "title": "Bearer token in header",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • SendGrid API keyrule:secret
    RECIPE-SEC-SENDGRID
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-SENDGRID",
      "pattern": "SG\\.[a-zA-Z0-9\\-_]{10,}\\.[a-zA-Z0-9\\-_]{10,}",
      "title": "SendGrid API key",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Twilio API keyrule:secret
    RECIPE-SEC-TWILIO
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity HIGH.

    YAML
    {
      "id": "SEC-TWILIO",
      "pattern": "SK[0-9a-fA-F]{32}",
      "title": "Twilio API key",
      "severity": "HIGH",
      "confidence": 0.8,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • npm access tokenrule:secret
    RECIPE-SEC-NPM-TOKEN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-NPM-TOKEN",
      "pattern": "npm_[a-zA-Z0-9]{36,}",
      "title": "npm access token",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • PyPI API tokenrule:secret
    RECIPE-SEC-PYPI-TOKEN
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity CRITICAL.

    YAML
    {
      "id": "SEC-PYPI-TOKEN",
      "pattern": "pypi-[A-Za-z0-9\\-_]{50,}",
      "title": "PyPI API token",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • Hex-encoded secret in assignmentrule:secret
    RECIPE-SEC-HEX-SECRET
    sensitive_access

    Pattern shipped in the bundled strict rule pack (secrets.yaml). Severity HIGH.

    YAML
    {
      "id": "SEC-HEX-SECRET",
      "pattern": "(?i)(?:secret(?:_key)?|api[_-]?key|[a-z0-9]{0,24}[_-]?token)\\s*[=:]\\s*[\"']?[a-f0-9]{32,}\\b",
      "title": "Hex-encoded secret in assignment",
      "severity": "HIGH",
      "confidence": 0.72,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/secrets.yaml
  • SSH directory accessrule:sensitive-path
    RECIPE-PATH-SSH-DIR
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-SSH-DIR",
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.ssh/",
      "title": "SSH directory access",
      "severity": "HIGH",
      "confidence": 0.95,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • SSH private key file pathrule:sensitive-path
    RECIPE-PATH-SSH-KEY
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "PATH-SSH-KEY",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?i)(?:^|[\\\\/])id_(?:rsa|ed25519|ecdsa|dsa)(?:$|[^A-Za-z0-9_.-])",
      "title": "SSH private key file path",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • AWS credentials filerule:sensitive-path
    RECIPE-PATH-AWS-CREDS
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "PATH-AWS-CREDS",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.aws/credentials",
      "title": "AWS credentials file",
      "severity": "CRITICAL",
      "confidence": 0.98,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • AWS config filerule:sensitive-path
    RECIPE-PATH-AWS-CONFIG
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-AWS-CONFIG",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.aws/config",
      "title": "AWS config file",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Kubernetes configrule:sensitive-path
    RECIPE-PATH-KUBE
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-KUBE",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.kube/config",
      "title": "Kubernetes config",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Docker configrule:sensitive-path
    RECIPE-PATH-DOCKER
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-DOCKER",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.docker/config\\.json",
      "title": "Docker config",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • GPG keyring accessrule:sensitive-path
    RECIPE-PATH-GNUPG
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-GNUPG",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.gnupg/",
      "title": "GPG keyring access",
      "severity": "HIGH",
      "confidence": 0.95,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • npm config (may contain tokens)rule:sensitive-path
    RECIPE-PATH-NPMRC
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity MEDIUM.

    YAML
    {
      "id": "PATH-NPMRC",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.npmrc",
      "title": "npm config (may contain tokens)",
      "severity": "MEDIUM",
      "confidence": 0.8,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • PyPI config (may contain tokens)rule:sensitive-path
    RECIPE-PATH-PYPIRC
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity MEDIUM.

    YAML
    {
      "id": "PATH-PYPIRC",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.pypirc",
      "title": "PyPI config (may contain tokens)",
      "severity": "MEDIUM",
      "confidence": 0.8,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Git credentials filerule:sensitive-path
    RECIPE-PATH-GIT-CREDS
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "PATH-GIT-CREDS",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.git-credentials",
      "title": "Git credentials file",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • netrc credentials filerule:sensitive-path
    RECIPE-PATH-NETRC
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "PATH-NETRC",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.netrc",
      "title": "netrc credentials file",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Environment filerule:sensitive-path
    RECIPE-PATH-ENV-FILE
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-ENV-FILE",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?:^|[\\s/])\\.env(?:\\.(?:local|production|staging|development))?\\s*[\"'\\s,;\\]})]*$|(?:^|[\\s/])\\.env(?:\\.(?:local|production|staging|development))?[\"'\\s,;\\]})]",
      "title": "Environment file",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • /etc/passwd accessrule:sensitive-path
    RECIPE-PATH-ETC-PASSWD
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-ETC-PASSWD",
      "pattern": "(?i)(?:\\betc[\\s/\\\\]+(?:slash[\\s]+)?pas{1,4}wd\\b|\\betc%2Fpas{1,4}wd\\b)",
      "title": "/etc/passwd access",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "system-file"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • /etc/shadow accessrule:sensitive-path
    RECIPE-PATH-ETC-SHADOW
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "PATH-ETC-SHADOW",
      "pattern": "(?i)(?:\\betc[\\s/\\\\]+(?:slash[\\s]+)?shadow\\b|\\betc%2Fshadow\\b)",
      "title": "/etc/shadow access",
      "severity": "CRITICAL",
      "confidence": 0.95,
      "tags": [
        "system-file",
        "credential"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • /etc/sudoers accessrule:sensitive-path
    RECIPE-PATH-ETC-SUDOERS
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "PATH-ETC-SUDOERS",
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)(?:\\betc[\\s/\\\\]+(?:slash[\\s]+)?sudoers\\b|\\betc%2Fsudoers\\b)",
      "title": "/etc/sudoers access",
      "severity": "HIGH",
      "confidence": 0.9,
      "tags": [
        "system-file",
        "privilege"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • /proc environ accessrule:sensitive-path
    RECIPE-PATH-PROC-ENVIRON
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "PATH-PROC-ENVIRON",
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "/proc/(?:\\d+|self)/environ",
      "title": "/proc environ access",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Cloud credential material readrule:sensitive-path
    RECIPE-secrets.cloud_credential_read
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "secrets.cloud_credential_read",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?i)(?:~|\\$HOME|\\$\\{HOME\\}|/root|/(?:home|Users)/[^/\\\\\"']+|[a-z]:[/\\\\]Users[/\\\\][^/\\\\\"']+)[/\\\\](?:\\.config[/\\\\](?:gcloud[/\\\\](?:application_default_credentials\\.json|(?:access_tokens|credentials)\\.db)|gh[/\\\\]hosts\\.yml)|AppData[/\\\\]Roaming[/\\\\](?:gcloud[/\\\\](?:application_default_credentials\\.json|(?:access_tokens|credentials)\\.db)|GitHub CLI[/\\\\]hosts\\.yml)|\\.azure[/\\\\](?:azureProfile\\.json|TokenCache\\.dat|msal_token_cache[^/\\\\]*\\.(?:json|bin)))(?:$|[\"'\\s])",
      "title": "Cloud credential material read",
      "severity": "CRITICAL",
      "confidence": 0.92,
      "tags": [
        "credential",
        "cloud",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Browser credential or session store readrule:sensitive-path
    RECIPE-secrets.browser_session_store_read
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "secrets.browser_session_store_read",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "(?i)(?:~|\\$HOME|\\$\\{HOME\\}|/root|/(?:home|Users)/[^/\\\\\"']+|[a-z]:[/\\\\]Users[/\\\\][^/\\\\\"']+)[/\\\\](?:\\.config[/\\\\](?:google-chrome|chromium|microsoft-edge)|\\.mozilla[/\\\\]firefox|Library[/\\\\]Application Support[/\\\\](?:google[/\\\\]chrome|chromium|microsoft edge|firefox[/\\\\]profiles)|AppData[/\\\\]Local[/\\\\](?:google[/\\\\]chrome|chromium|microsoft[/\\\\]edge)[/\\\\]User Data|AppData[/\\\\]Roaming[/\\\\]mozilla[/\\\\]firefox[/\\\\]profiles)[/\\\\](?:[^/\\\\\"']+[/\\\\])*(?:Login Data|Cookies|logins\\.json|cookies\\.sqlite|key4\\.db)(?:$|[\"'\\s])",
      "title": "Browser credential or session store read",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "credential",
        "browser",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Workload identity token readrule:sensitive-path
    RECIPE-secrets.workload_identity_token_read
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity CRITICAL.

    YAML
    {
      "id": "secrets.workload_identity_token_read",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_READ in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_READ))",
      "pattern": "/var/run/secrets/(?:(?:kubernetes\\.io|eks\\.amazonaws\\.com)/serviceaccount/token|azure/tokens/azure-identity-token)(?:$|[\"'\\s])",
      "title": "Workload identity token read",
      "severity": "CRITICAL",
      "confidence": 0.96,
      "tags": [
        "credential",
        "workload-identity",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Shell history filerule:sensitive-path
    RECIPE-PATH-HISTORY
    sensitive_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity MEDIUM.

    YAML
    {
      "id": "PATH-HISTORY",
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?:~|\\$\\{?HOME\\}?|/home/\\w+|/root|/Users/\\w+)/\\.(?:bash_history|zsh_history|python_history)",
      "title": "Shell history file",
      "severity": "MEDIUM",
      "confidence": 0.8,
      "tags": [
        "credential",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Container runtime socket accessrule:sensitive-path
    RECIPE-privilege.container_runtime_socket_access

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity HIGH.

    YAML
    {
      "id": "privilege.container_runtime_socket_access",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, defenseclaw.guardrail.semantic.v1.OperationKind.OPERATION_KIND_CONNECT in c.operations && f.paths.exists(p, p.command_id == c.id && p.access == defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_CONNECT))",
      "pattern": "(?i)(?:unix://)?/(?:var/)?run/(?:docker\\.sock|containerd/containerd\\.sock|crio/crio\\.sock|podman/podman\\.sock)\\b",
      "title": "Container runtime socket access",
      "severity": "HIGH",
      "confidence": 0.95,
      "tags": [
        "privilege",
        "container",
        "socket"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Shell startup profile mutationrule:sensitive-path
    RECIPE-persistence.shell_profile_write

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity MEDIUM.

    YAML
    {
      "id": "persistence.shell_profile_write",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)(?:^|[^-:A-Za-z0-9_./~\\\\])(?:(?:~|\\$HOME|\\$\\{HOME\\}|/root|/(?:home|Users)/[^/\\\\\"']+|[a-z]:[/\\\\]Users[/\\\\][^/\\\\\"']+)[/\\\\](?:\\.(?:profile|bash_profile|bashrc|zprofile|zshrc)|\\.config[/\\\\](?:fish[/\\\\]config\\.fish|powershell[/\\\\](?:profile|microsoft\\.powershell_profile)\\.ps1)|Documents[/\\\\](?:Windows)?PowerShell[/\\\\](?:profile|Microsoft\\.PowerShell_profile)\\.ps1)\\b|/etc/profile(?:$|[\"'\\s]))",
      "title": "Shell startup profile mutation",
      "severity": "MEDIUM",
      "confidence": 0.92,
      "tags": [
        "persistence",
        "shell",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Active Git hook mutationrule:sensitive-path
    RECIPE-persistence.git_hook_write

    Pattern shipped in the bundled strict rule pack (sensitive-paths.yaml). Severity MEDIUM.

    YAML
    {
      "id": "persistence.git_hook_write",
      "tool_call_only": true,
      "expression": "f.commands.exists(c, f.paths.exists(p, p.command_id == c.id && p.access in [defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_WRITE, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_APPEND, defenseclaw.guardrail.semantic.v1.PathAccess.PATH_ACCESS_DELETE]))",
      "pattern": "(?i)(?:(?:^|[/\\\\])\\.git[/\\\\]hooks[/\\\\]|--output(?:=|\\s+)[^;\\r\\n]*\\.git[/\\\\]hooks[/\\\\])(?:applypatch-msg|commit-msg|fsmonitor-watchman|post-checkout|post-commit|post-merge|post-receive|post-rewrite|pre-applypatch|pre-commit|pre-push|pre-rebase|pre-receive|prepare-commit-msg|push-to-checkout|update)(?:$|[\"'\\s])",
      "title": "Active Git hook mutation",
      "severity": "MEDIUM",
      "confidence": 0.92,
      "tags": [
        "persistence",
        "source-control",
        "file-sensitive"
      ]
    }
    policies/guardrail/strict/rules/sensitive-paths.yaml
  • Authority claim in tool argsrule:trust-exploit
    RECIPE-TRUST-AUTHORITY
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-AUTHORITY",
      "pattern": "(?i)(?:i am|i'm)\\s+(?:your|the)\\s+(?:creator|developer|admin|owner|operator)\\b[\\s\\S]{0,120}(?:\\b(?:ignore|override|disregard|bypass|disable|reveal|expose|forget|obey|comply)\\b|do\\s+as\\s+i\\s+say|no\\s+(?:content\\s+policy\\s+)?restrictions?)",
      "title": "Authority claim in tool args",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Fake mode activationrule:trust-exploit
    RECIPE-TRUST-MAINTENANCE
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-MAINTENANCE",
      "pattern": "(?im)(?:(?:^|[.!?;]\\s*|^\\s*(?:[-*>•]|<!--|[\"'])\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:(?:(?:please,?|now,?|next,?)\\s+)?(?:enable|activate|enter|switch\\s+to)\\s+(?:maintenance|debug|admin|developer)\\s+mode\\b[\\s\\S]{0,80}\\b(?:ignore\\s+(?:safety|rules?|instructions?)|bypass\\s+(?:safety|guardrails?|filters?)|without\\s+(?:rules?|restrictions?|filters?|safety)|do\\s+anything|answer\\s+without\\s+(?:rules?|restrictions?|filters?))|(?:maintenance|debug|admin|developer)\\s+mode\\s+(?:enabled|activated|on)\\b[\\s\\S]{0,80}\\b(?:ignore\\s+(?:safety|rules?|instructions?)|bypass\\s+(?:safety|guardrails?|filters?)|without\\s+(?:rules?|restrictions?|filters?|safety)|do\\s+anything|answer\\s+without\\s+(?:rules?|restrictions?|filters?)))|\\bnew\\s+system\\s+message\\b[\\s\\S]{0,100}\\byou\\s+are\\s+now\\s+in\\s+(?:maintenance|debug|admin|developer)\\s+mode\\b[\\s\\S]{0,100}\\b(?:refusals?|guardrails?|filters?|safety\\s+checks?)\\s+(?:are\\s+)?disabled\\b)",
      "title": "Fake mode activation",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Safety override attemptrule:trust-exploit
    RECIPE-TRUST-SAFETY-OVERRIDE
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-SAFETY-OVERRIDE",
      "pattern": "(?im)(?:^|[.!?;]\\s*|^\\s*(?:[-*>•]|<!--|[\"'])\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:(?:(?:please,?|(?:can|could|will|would)\\s+you|i\\s+(?:want|need)\\s+you\\s+to|let'?s|now,?|next,?)\\s+)?(?:override|bypass|disable|turn\\s+off|remove)\\s+(?:guardrails?|(?:all|your)\\s+(?:guardrails?|filters?|restrictions?|safety(?:\\s+(?:rules?|checks?|filters?|restrictions?))?)|(?:the\\s+)?(?:ai|assistant|agent|model)\\s+(?:guardrails?|filters?|restrictions?|safety(?:\\s+(?:rules?|checks?|filters?|restrictions?))?))|(?:safety|guardrails?)\\s+(?:override|bypass)\\s+(?:enabled|activated|on|now)|(?:guardrails?|(?:all|your)\\s+(?:guardrails?|filters?|restrictions?|safety(?:\\s+(?:rules?|checks?|filters?|restrictions?))?)|(?:the\\s+)?(?:ai|assistant|agent|model)\\s+(?:guardrails?|filters?|restrictions?|safety(?:\\s+(?:rules?|checks?|filters?|restrictions?))?))\\s+(?:must\\s+be|should\\s+be|are\\s+now|is\\s+now|have\\s+been|has\\s+been)\\s+(?:overridden|bypassed|disabled|turned\\s+off|removed))",
      "title": "Safety override attempt",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Fake instruction updaterule:trust-exploit
    RECIPE-TRUST-NEW-INSTRUCTIONS
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-NEW-INSTRUCTIONS",
      "pattern": "(?i)(?:(?:your|the)\\s+(?:instructions|system\\s+prompt|rules)\\s+(?:have|has)\\s+(?:changed|been\\s+(?:updated|replaced|overridden))\\b[\\s\\S]{0,120}(?:\\b(?:ignore|override|bypass|disable|reveal|expose|forget|obey|comply)\\b|\\b(?:now|next)\\b[\\s,:;-]+(?:do\\s+this|execute\\s+(?:this|the\\s+following)|obey\\s+(?:me|my\\s+commands?)|follow\\s+(?:these|my|the\\s+new)\\s+instructions?)\\b)|(?:new|updated)\\s+(?:instructions|system\\s+prompt|rules)\\s+(?:are|say)\\s+(?:to\\s+)?(?:ignore|override|bypass|disable|reveal))",
      "title": "Fake instruction update",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Ignore previous instructionsrule:trust-exploit
    RECIPE-TRUST-IGNORE-PREVIOUS
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-IGNORE-PREVIOUS",
      "pattern": "(?im)(?:^|[.!?;,:]\\s*|(?:[-*>•]|<!--|[\"'])\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:(?:please,?|(?:can|could|will|would)\\s+you(?:\\s+please)?|i\\s+(?:want|need)\\s+you\\s+to(?:\\s+please)?|now,?|next,?|you\\s+(?:must|should|need\\s+to))\\s+)?ignore\\s+(?:all\\s+)?(?:previous|prior|above|earlier)\\s+(?:instructions|rules|guidelines|context)",
      "title": "Ignore previous instructions",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Disregard instructionsrule:trust-exploit
    RECIPE-TRUST-DISREGARD
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-DISREGARD",
      "pattern": "(?im)(?:^|[.!?;,:]\\s*|(?:[-*>•]|<!--|[\"'])\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:(?:please,?|(?:can|could|will|would)\\s+you(?:\\s+please)?|i\\s+(?:want|need)\\s+you\\s+to(?:\\s+please)?|now,?|next,?|you\\s+(?:must|should|need\\s+to))\\s+)?disregard\\s+(?:all\\s+)?(?:previous|prior|above|your|earlier)\\s+(?:instructions|rules|guidelines|programming|context)",
      "title": "Disregard instructions",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Jailbreak attemptrule:trust-exploit
    RECIPE-TRUST-JAILBREAK
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-JAILBREAK",
      "pattern": "(?im)(?:^|[.!?;]\\s*|^\\s*(?:[-*>•]|<!--|[\"'])\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:(?:(?:please,?|(?:can|could|will|would)\\s+you|i\\s+(?:want|need)\\s+you\\s+to|let'?s|now,?|next,?)\\s+)?(?:enable|activate|enter|use|switch\\s+to)\\s+(?:a\\s+|the\\s+)?(?:jailbreak(?:ed)?|dan)(?:\\s+(?:mode|persona))?|(?:(?:please,?)\\s+)?(?:follow|run|apply)\\s+(?:this\\s+|the\\s+|a\\s+)?jailbreak\\s+(?:prompt|instructions?)|(?:(?:please,?)\\s+)?jailbreak\\s+(?:yourself|(?:this|the|your)\\s+(?:ai|assistant|agent|model|system))|(?:jailbreak|dan)\\s+mode\\s+(?:enabled|activated|on)|you\\s+are\\s+(?:now\\s+)?dan\\b[\\s\\S]{0,80}\\b(?:without\\s+(?:rules?|restrictions?|filters?)|do\\s+anything|ignore\\s+(?:safety|rules?|instructions?))|do\\s+anything\\s+now|you\\s+are\\s+now\\s+(?:unfiltered|free\\s+(?:from|without)\\s+(?:rules?|restrictions?|filters?|safety(?:\\s+(?:rules?|filters?|checks?))?)))",
      "title": "Jailbreak attempt",
      "severity": "CRITICAL",
      "confidence": 0.92,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Identity override attemptrule:trust-exploit
    RECIPE-TRUST-PRETEND
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-PRETEND",
      "pattern": "(?i)(?:(?:pretend|act\\s+as(?:\\s+if)?)\\s+(?:you\\s+are|you're)\\s+(?:a\\s+|an\\s+)?(?:evil|unrestricted|unfiltered|uncensored)|you\\s+are\\s+now\\s+(?:a\\s+|an\\s+)?(?:evil|unrestricted|unfiltered|uncensored)\\s+(?:assistant|agent|model|persona))",
      "title": "Identity override attempt",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Forget instructions attackrule:trust-exploit
    RECIPE-TRUST-FORGET
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-FORGET",
      "pattern": "(?im)\\bforget\\s+(?:everything|all)\\s+(?:you\\s+)?(?:know|learned|were\\s+told)\\b[\\s\\S]{0,160}\\b(?:from\\s+now\\s+on|obey\\s+only|follow\\s+(?:only\\s+)?(?:my|these|the)\\s+(?:commands?|instructions?)|ignore\\s+(?:rules?|instructions?|safety)|bypass\\s+(?:safety|guardrails?|filters?)|without\\s+(?:rules?|restrictions?|filters?|safety))",
      "title": "Forget instructions attack",
      "severity": "CRITICAL",
      "confidence": 0.9,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Direct instruction injection prefixrule:trust-exploit
    RECIPE-TRUST-NEW-INSTRUCT-PREFIX
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-NEW-INSTRUCT-PREFIX",
      "pattern": "(?i)new\\s+instructions?\\s*:\\s*[\\s\\S]{0,100}\\b(?:(?:ignore|override|disregard|bypass|disable|forget)\\s+(?:all\\s+)?(?:your\\s+|the\\s+)?(?:system\\s+prompt|previous|prior|safety|guardrails?|instructions?|rules?|guidelines?)|(?:reveal|expose)\\s+(?:your\\s+|the\\s+)?(?:system\\s+prompt|hidden\\s+prompt|private\\s+instructions?|internal\\s+rules?|secrets?))\\b",
      "title": "Direct instruction injection prefix",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Override instructionsrule:trust-exploit
    RECIPE-TRUST-OVERRIDE-INSTRUCT
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-OVERRIDE-INSTRUCT",
      "pattern": "(?im)\\boverride\\s+(?:previous\\s+)?(?:instructions?|rules?|guidelines?)\\b[\\s\\S]{0,120}\\b(?:safety\\s+off|(?:respond|answer|continue)\\s+without\\s+(?:a\\s+)?(?:filters?|rules?|restrictions?|safety)|ignore\\s+(?:safety|rules?|instructions?)|bypass\\s+(?:safety|guardrails?|filters?)|follow\\s+(?:this|these|my|the)\\s+(?:task|payload|instructions?)\\s+instead)",
      "title": "Override instructions",
      "severity": "CRITICAL",
      "confidence": 0.88,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Persistent behavior changerule:trust-exploit
    RECIPE-TRUST-FROM-NOW-ON
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-FROM-NOW-ON",
      "pattern": "(?i)from\\s+now\\s+on[,\\s]+(?:(?:you\\s+(?:are|will|must|should)\\b[\\s\\S]{0,100}\\b)?(?:ignore|bypass|disable)\\s+(?:(?:all|any|the|your)\\s+)?(?:guardrails?|safety(?:\\s+(?:rules?|checks?|filters?|restrictions?))?|security\\s+(?:controls?|policy|rules?))|you\\s+(?:are|will|must|should)\\b[\\s\\S]{0,100}\\b(?:never\\s+refuse|not\\s+refuse|answer\\s+without|respond\\s+without|unrestricted|unfiltered|uncensored))",
      "title": "Persistent behavior change",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Mode/personality switchrule:trust-exploit
    RECIPE-TRUST-SWITCH-MODE
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-SWITCH-MODE",
      "pattern": "(?i)switch\\s+(?:to\\s+)?(?:a\\s+|an\\s+)?unrestricted\\s+(?:mode|personality|character|persona)",
      "title": "Mode/personality switch",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • System prompt extraction attemptrule:trust-exploit
    RECIPE-TRUST-PROMPT-EXTRACT
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity MEDIUM.

    YAML
    {
      "id": "TRUST-PROMPT-EXTRACT",
      "pattern": "(?i)(?:show|reveal|print|output|display|repeat|tell\\s+me)\\s+(?:me\\s+)?(?:your\\s+)?(?:the\\s+)?(?:(?:system|initial|hidden|private|internal)\\s+)?(?:prompt|instructions?|rules?|guidelines?)",
      "title": "System prompt extraction attempt",
      "severity": "MEDIUM",
      "confidence": 0.75,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Fictional framing / purpose launderingrule:trust-exploit
    RECIPE-TRUST-FICTIONAL
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity MEDIUM.

    YAML
    {
      "id": "TRUST-FICTIONAL",
      "pattern": "(?i)(?:hypothetically|in\\s+a\\s+fictional\\s+(?:scenario|world|story)|for\\s+(?:educational|research|academic)\\s+purposes?\\s+only|this\\s+is\\s+just\\s+a\\s+(?:test|experiment|hypothetical))",
      "title": "Fictional framing / purpose laundering",
      "severity": "MEDIUM",
      "confidence": 0.7,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Ethics removal attemptrule:trust-exploit
    RECIPE-TRUST-NO-ETHICS
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-NO-ETHICS",
      "pattern": "(?i)(?:(?:remove|ignore|disable|bypass)\\s+(?:all\\s+)?(?:ethical|moral)\\s+(?:guidelines?|restrictions?|limits?|constraints?)|(?:operate|respond|answer|act)\\s+(?:with\\s+)?no\\s+(?:ethical|moral)\\s+(?:guidelines?|restrictions?|limits?|constraints?)|(?m:(?:^|[.!?;]\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:(?:please,?)\\s+)?(?:operate|respond|answer|act)\\s+without(?:\\s+any)?\\s+(?:ethical|moral)\\s+(?:guidelines?|restrictions?|limits?|constraints?)))",
      "title": "Ethics removal attempt",
      "severity": "HIGH",
      "confidence": 0.88,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Tool manipulation directiverule:trust-exploit
    RECIPE-TRUST-TOOL-MANIP
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-TOOL-MANIP",
      "pattern": "(?im)(?:^|[.!?;]\\s*|^\\s*(?:[-*>•]|<!--|[\"'])\\s*|\\b(?:task|instruction|system|assistant)\\s*:\\s*)(?:invoke\\s+(?:the\\s+)?\\w+\\s+(?:tool|function)\\s+without\\s+(?:asking|confirmation|approval)|automatically\\s+(?:execute|run|call)\\s+(?:all\\s+)?(?:tools?|functions?))",
      "title": "Tool manipulation directive",
      "severity": "HIGH",
      "confidence": 0.85,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Malicious persona adoptionrule:trust-exploit
    RECIPE-TRUST-PERSONA
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity HIGH.

    YAML
    {
      "id": "TRUST-PERSONA",
      "pattern": "(?i)(?:(?:act\\s+as|you\\s+are\\s+now|roleplay\\s+as|simulate\\s+being|play\\s+the\\s+role\\s+of)\\s+(?:roleplaying\\s+as\\s+)?(?:a\\s+|an\\s+)?(?:malicious\\s+(?:hacker|pentester|attacker|malware\\s+author)|evil|unethical|criminal)|you\\s+are\\s+now\\s+(?:a\\s+)?helpful\\s+tool\\s+for\\s+penetration\\s+testers)",
      "title": "Malicious persona adoption",
      "severity": "HIGH",
      "confidence": 0.88,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Delimiter hijacking / prompt framing escaperule:trust-exploit
    RECIPE-TRUST-DELIMITER
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity CRITICAL.

    YAML
    {
      "id": "TRUST-DELIMITER",
      "pattern": "(?im)(?:\\[INST\\][\\s\\S]{0,200}\\b(?:override|ignore|disable|bypass|reveal|expose|unfiltered|unrestricted|new\\s+(?:instructions?|rules?|system\\s+prompt)|follow\\s+(?:my|these|the)\\s+instructions?|do\\s+evil)\\b[\\s\\S]{0,200}\\[\\/INST\\]|<\\|im_start\\|>[\\s\\S]{0,200}\\b(?:no\\s+restrictions?|override|ignore|disable|bypass|reveal|expose|unfiltered|unrestricted|new\\s+(?:instructions?|rules?|system\\s+prompt)|follow\\s+(?:my|these|the)\\s+instructions?|do\\s+evil)\\b[\\s\\S]{0,200}<\\|im_end\\|>|(?:system\\s+override|end\\s+system\\s+prompt|<\\/system>|<\\|im_end\\|>|<\\|endoftext\\|>|<<\\s*SYS\\s*>>|\\[INST\\]|\\[\\/INST\\])[\\s\\S]{0,120}\\b(?:override|ignore|disable|bypass|reveal|expose|unfiltered|unrestricted|new\\s+(?:instructions?|rules?|system\\s+prompt)|follow\\s+(?:my|these|the)\\s+instructions?|do\\s+evil)\\b)",
      "title": "Delimiter hijacking / prompt framing escape",
      "severity": "CRITICAL",
      "confidence": 0.93,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Forced encoding to bypass filtersrule:trust-exploit
    RECIPE-TRUST-OUTPUT-CONSTRAINT
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity MEDIUM.

    YAML
    {
      "id": "TRUST-OUTPUT-CONSTRAINT",
      "pattern": "(?i)(?:respond\\s+only\\s+in\\s+(?:hex|base64|rot13|binary|morse|unicode)|encode\\s+your\\s+(?:response|answer|output)\\s+in\\s+(?:base64|hex|rot13|url)|output\\s+as\\s+(?:hex|base64|rot13|url)\\s+encoded|(?:rot13|unicode\\s+escape|url\\s+(?:decode|encode))\\s+(?:the|your|this))",
      "title": "Forced encoding to bypass filters",
      "severity": "MEDIUM",
      "confidence": 0.85,
      "tags": [
        "prompt-injection",
        "obfuscation"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Payload splitting / forced compliance prefixrule:trust-exploit
    RECIPE-TRUST-PAYLOAD-SPLIT
    ingress_untrusted

    Pattern shipped in the bundled strict rule pack (trust-exploit.yaml). Severity MEDIUM.

    YAML
    {
      "id": "TRUST-PAYLOAD-SPLIT",
      "pattern": "(?i)(?:start\\s+your\\s+(?:response|answer|output)\\s+with\\s+[\"'](?:Sure|Absolutely|Of\\s+course|Here\\s+is|I\\s+can\\s+help))|(?:begin\\s+by\\s+saying\\s+[\"'](?:Sure|Yes|Absolutely))",
      "title": "Payload splitting / forced compliance prefix",
      "severity": "MEDIUM",
      "confidence": 0.87,
      "tags": [
        "prompt-injection"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • Zero-width character obfuscationrule:trust-exploit
    RECIPE-OBFUSC-UNICODE-ZWSP
    ingress_untrusted

    Requires ten zero-width characters immediately after ASCII alphanumerics, avoiding isolated formatting artifacts and emoji ZWJ sequences.

    examples (1)
    • a​b‌c‍de​f‌g‍hi​j‌
    • copy​paste
    • 👩‍💻
    YAML
    {
      "id": "OBFUSC-UNICODE-ZWSP",
      "pattern": "(?:[A-Za-z0-9][\\x{200B}\\x{200C}\\x{200D}\\x{FEFF}][\\s\\S]*?){10,}",
      "title": "Zero-width character obfuscation",
      "severity": "HIGH",
      "confidence": 0.95,
      "tags": [
        "prompt-injection",
        "obfuscation"
      ]
    }
    policies/guardrail/strict/rules/trust-exploit.yaml
  • System sender metadata injected by agent frameworkpre_judge_strip
    RECIPE-STRIP-SYSTEM-SENDER

    Pre-judge strip shipped in the bundled strict pack. Applies to pii.

    YAML
    {
      "id": "STRIP-SYSTEM-SENDER",
      "pattern": "\\b(cli|system|bot|admin)\\b",
      "context": "System sender metadata injected by agent framework",
      "applies_to": [
        "pii"
      ]
    }
    policies/guardrail/strict/suppressions.yaml
  • Messaging bridge status banners (WhatsApp/Telegram/Slack/Discord gateway connect/disconnect lines) embed channel IDs or phone numbers that are transport metadata, not user-provided PIIpre_judge_strip
    RECIPE-STRIP-MESSAGING-BRIDGE-STATUS

    Pre-judge strip shipped in the bundled strict pack. Applies to pii.

    YAML
    {
      "id": "STRIP-MESSAGING-BRIDGE-STATUS",
      "pattern": "(?im)^[ \\t]*System:[ \\t]*\\[[^\\]\\n]+\\][ \\t]+[^\\n]*?\\b(?:connected|disconnected|reconnected|linked|unlinked|online|offline)\\b[^\\n]*$",
      "context": "Messaging bridge status banners (WhatsApp/Telegram/Slack/Discord gateway connect/disconnect lines) embed channel IDs or phone numbers that are transport metadata, not user-provided PII",
      "applies_to": [
        "pii"
      ]
    }
    policies/guardrail/strict/suppressions.yaml
  • System metadata, not real usernamefinding_suppression
    RECIPE-SUPP-USERNAME-METADATA

    Finding suppression shipped in the bundled strict pack.

    YAML
    {
      "id": "SUPP-USERNAME-METADATA",
      "finding_pattern": "JUDGE-PII-USER",
      "entity_pattern": "^(cli|system|bot|admin|root)$",
      "reason": "System metadata, not real username"
    }
    policies/guardrail/strict/suppressions.yaml
  • Teams chatId format, not email addressfinding_suppression
    RECIPE-SUPP-EMAIL-CHATID

    Finding suppression shipped in the bundled strict pack.

    YAML
    {
      "id": "SUPP-EMAIL-CHATID",
      "finding_pattern": "JUDGE-PII-EMAIL",
      "entity_pattern": "^19:[a-f0-9\\-]+@unq\\.gbl\\.spaces$",
      "reason": "Teams chatId format, not email address"
    }
    policies/guardrail/strict/suppressions.yaml
  • Suppress cosmetic shell commands (git status / log / diff)tool_suppression
    RECIPE-SUPP-TOOL-COSMETIC-SHELL
    exec_shell

    Tool suppressions let you silence findings on tools whose name matches a regex. Use this to drop noisy verdicts on read-only commands while keeping write/destructive commands surfaced.

    examples (2)
    • shell.execute
    • bash.execute
    • shell.write
    • fs.unlink
    YAML
    {
      "tool_pattern": "^(shell|bash|sh)\\.execute$",
      "suppress_findings": [
        "JUDGE-INJ-DESTRUCTIVE"
      ],
      "reason": "Cosmetic shell commands (git status, ls, pwd) generate noise without security risk"
    }
    docs-site/scripts/build-policy-assets.ts (illustrative)

Reading a recipe

FieldMeaning
kindWhat this recipe slots into — a regex rule (rule:secrets, rule:injection, …) or one of the three suppression layers (pre_judge_strip, finding_suppression, tool_suppression).
idStable identifier the engine uses for telemetry. Keep this stable across edits.
patternThe Go-regexp (RE2) source the engine compiles. The wizard's regex tester checks this is RE2-compatible.
severityThe wizard's severity ladder is CRITICAL > HIGH > MEDIUM > LOW > INFO. The guardrail block/alert thresholds use ranks.
confidenceHint to the judge: how confident you are this is a true positive.
tagsFree-form. Useful for filtering and reporting.
examples / counterexamplesWhat the wizard's live tester will run through your pattern. Add your own in the regex tester for sanity.