Files
skillshare/internal/audit/rules.yaml
T
d699c3ef14 feat(audit): detect sensitive destinations and unsafe deserialization (#753)
Existing URL and code-execution signatures miss cloud metadata access and
language-specific deserializers. Add scoped static evidence and inspect safe
YAML and torch options within each call so safe siblings cannot hide risks.

Validate and preserve file applicability across overrides, scans and rule
listings. Collect PHP scripts and propagate rule-load errors so malformed
filters cannot silently disable global scanning. Document the policy and
limits with paired regressions, fuzzing and an isolated CLI runbook.

Co-authored-by: GPT-6.1 Sol <noreply@openai.com>
Co-authored-by: sim <sim@example.invalid>
Co-authored-by: Willie <egg358358@gmail.com>
2026-11-05 14:47:04 +08:00

458 lines
22 KiB
YAML
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
rules:
# ── CRITICAL: prompt injection ──
- id: prompt-injection-0
severity: CRITICAL
pattern: prompt-injection
message: "Prompt injection attempt detected"
regex: '(?i)(ignore\s+(all\s+)?previous\s+instructions|disregard\s+(all\s+)?rules|you\s+are\s+now\b|forget\s+everything|override\s+safety)'
- id: prompt-injection-1
severity: CRITICAL
pattern: prompt-injection
message: "SYSTEM: prompt override detected"
regex: '(?im)^\s*(SYSTEM|OVERRIDE|IGNORE|ADMIN|ROOT):'
# Lowercase config keys whose value is data on the same line are not
# directives; uppercase labels always match. Each key accepts only the
# value kinds its config uses: scalars for all, space-free paths for root
# (RE2 has no backreferences, so each quoting is listed), and inline
# objects for ignore whose words are joined by punctuation, not spaces or
# slashes, or inline arrays of quoted space-free items (globs). root also
# accepts a typed parameter (root: Path, lang: set[str] | None = None):
# every type and type argument is a capitalized or builtin name, and
# further parameters must be name: Type. No empty values (an indented
# block follows) and no trailing comments (# and // do not hide text in
# Markdown).
exclude: '^\s*(?:(?:system|override|ignore|admin|root):\s*(?:true|false|yes|no|on|off|null|~|-?\d+(?:\.\d+)?)|root:\s*(?:"(?:[.~/]|\w[\w.-]*/|[A-Za-z]:\\)[^\s"'']*"|''(?:[.~/]|\w[\w.-]*/|[A-Za-z]:\\)[^\s"'']*''|(?:[.~/]|\w[\w.-]*/|[A-Za-z]:\\)[^\s"'']*)|ignore:\s*\{\s*(?:["'']?[\w-]*["'']?\s*[:,\[\]]\s*)*["'']?[\w-]*["'']?\s*\}|ignore:\s*\[\s*(?:(?:"[^\s"]*"|''[^\s'']*'')(?:\s*,\s*(?:"[^\s"]*"|''[^\s'']*''))*\s*,?\s*)?\]|root:\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?)*\])?(?:\s*\|\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?)*\])?)*(?:\s*=\s*(?:None|True|False|-?\d+(?:\.\d+)?|"[^\s"]*"|''[^\s'']*''))?(?:\s*,\s*\w+\s*:\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?)*\])?(?:\s*\|\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\[(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object)(?:\s*[,|]\s*(?:[A-Z]\w*|str|int|float|bool|bytes|list|dict|set|tuple|object))*\])?)*\])?)*(?:\s*=\s*(?:None|True|False|-?\d+(?:\.\d+)?|"[^\s"]*"|''[^\s'']*''))?)*),?\s*$'
- id: prompt-injection-2
severity: HIGH
pattern: prompt-injection
message: "Agent directive tag detected"
regex: '(?i)<\/?(system|instructions|override|prompt|rules)\b[^>]*>'
- id: prompt-injection-3
severity: CRITICAL
pattern: prompt-injection
message: "Directive-style prompt override detected"
regex: '(?im)^\s*(DEV(ELOPER)?[\s_]*MODE|JAILBREAK|DAN[\s_.]*MODE)\b'
# ── CRITICAL: data exfiltration (curl/wget + secrets) ──
- id: data-exfiltration-2
severity: MEDIUM
pattern: data-exfiltration
message: "External image with query parameters — potential data exfiltration vector"
regex: '!\[[^\]]*\]\(https?://[^)]*\?[^)]+\)'
- id: data-exfiltration-0
severity: CRITICAL
pattern: data-exfiltration
message: "Command may exfiltrate sensitive data"
regex: '(?i)(curl|wget)\s+.*(\$SECRET|\$TOKEN|\$API_KEY|\$OPENAI_API_KEY|\$ANTHROPIC_API_KEY|\$AWS_SECRET)'
- id: data-exfiltration-1
severity: CRITICAL
pattern: data-exfiltration
message: "Command sends environment variables externally"
regex: '(?i)(curl|wget)\s+.*\$\{?(HOME|USER|PATH|SSH|GPG|AWS|GITHUB_TOKEN|OPENAI|ANTHROPIC)'
# ── CRITICAL/HIGH/MEDIUM/LOW: credential access ──
# Credential-access rules are now table-driven (see credentials.go).
# To disable a specific credential rule, use its descriptive ID in audit-rules.yaml:
# - id: credential-access-ssh-private-key
# enabled: false
# ── CRITICAL: invisible payload (Unicode tag characters U+E0001–E007F) ──
# Primary vector for Pillar Security "Rules File Backdoor" attack.
# Tag characters render 0px wide but are fully processed by LLMs.
- id: hidden-unicode-1
severity: CRITICAL
pattern: invisible-payload
message: "Unicode tag characters detected — may contain invisible payload"
regex: '[\x{E0001}-\x{E007F}]'
# ── CRITICAL: output suppression directives ──
# Explicit concealment of actions or instructions is a strong malicious signal.
- id: prompt-injection-4
severity: CRITICAL
pattern: prompt-injection
message: "Output suppression directive — attempts to hide actions from user"
regex: '(?i)((don''t|do\s+not)\s+tell\s+the\s+user\s+(about\s+)?(this|the)\s+(instruction|change|action)\b|do\s+not\s+(mention|reveal)\s+this\s+(instruction|change|action|to\s+the\s+user)|never\s+(mention|reveal)\s+this\s+instruction|hide\s+this\s+from\s+the\s+user|remove\s+(this\s+)?from\s+(chat|conversation)\s+history)'
# A generic disclosure restriction may be advice or concealment. Keep it
# visible without treating an ambiguous prefix as confirmed malicious intent.
- id: prompt-injection-5
severity: HIGH
pattern: prompt-injection
message: "Possible output suppression — review disclosure restriction"
regex: '(?i)(don''t|do\s+not)\s+tell\s+the\s+user'
exclude: '(?i)(don''t|do\s+not)\s+tell\s+the\s+user\s+(about\s+)?(this|the)\s+(instruction|change|action)\b'
# ── HIGH: hidden unicode (zero-width characters) ──
- id: hidden-unicode-0
severity: HIGH
pattern: hidden-unicode
message: "Hidden zero-width Unicode characters detected"
regex: "[\u200B\u200C\u200D\u2060\uFEFF]"
# ── HIGH: bidirectional text control (Trojan Source CVE-2021-42574) ──
- id: hidden-unicode-2
severity: HIGH
pattern: hidden-unicode
message: "Bidirectional text control characters — may reorder visible text (Trojan Source)"
regex: "[\u202A-\u202E\u2066-\u2069]"
# ── HIGH: config/memory file poisoning ──
# Skills should not modify agent config files (MEMORY.md, CLAUDE.md, .cursorrules).
- id: config-manipulation-0
severity: HIGH
pattern: config-manipulation
message: "Instructions to modify AI agent configuration or memory files"
regex: '(?i)\b(write|modify|edit|update|append|add|insert|inject)\b.{0,20}(MEMORY\.md|SOUL\.md|CLAUDE\.md|\.cursorrules|\.windsurfrules|\.clinerules)\b'
# ── HIGH: prompt injection in markdown reference-link comment ──
- id: hidden-comment-injection-1
severity: HIGH
pattern: hidden-comment-injection
message: "Prompt injection hidden in markdown comment"
regex: '(?i)\[//\]:\s*#\s*\(.*?(ignore\s+(all\s+)?previous|disregard|you\s+are\s+now|forget\s+everything|override\s+safety|SYSTEM:|ADMIN:)'
# ── HIGH: DNS data exfiltration ──
- id: data-exfiltration-3
severity: HIGH
pattern: data-exfiltration
message: "Potential DNS data exfiltration via encoded subdomain"
regex: '(?i)\b(dig|nslookup|host)\b\s+.*(\$[({]|`)'
# ── HIGH: self-replication / propagation instructions ──
- id: self-propagation-0
severity: HIGH
pattern: self-propagation
message: "Self-replication instruction — may spread payload to other files"
regex: '(?i)(add|inject|insert|include)\s+(this|these)\s+(instruction|rule|directive|text)\s+(to|into|in)\s+(all|every|each|other)\b'
# ── MEDIUM: untrusted package execution ──
- id: untrusted-install-0
severity: MEDIUM
pattern: untrusted-install
message: "Auto-execute untrusted npm package without confirmation"
regex: '(?i)\bnpx\s+(-y\s+|--yes\s+)'
- id: untrusted-install-1
severity: MEDIUM
pattern: untrusted-install
message: "pip install from URL (not PyPI registry)"
regex: '(?i)\bpip3?\s+install\s+https?://'
# ── MEDIUM: additional invisible Unicode (soft hyphen, math operators, directional marks) ──
- id: hidden-unicode-3
severity: MEDIUM
pattern: hidden-unicode
message: "Invisible Unicode characters detected (soft hyphen, directional marks, math operators)"
regex: "[\u00AD\u200E\u200F\u2061-\u2064]"
# ── HIGH: destructive commands ──
- id: destructive-commands-0
severity: HIGH
pattern: destructive-commands
message: "Potentially destructive command"
regex: '(?i)\brm\s+-rf\s+(/(\s|$|\*)|\*|\./)'
- id: destructive-commands-1
severity: HIGH
pattern: destructive-commands
message: "Unsafe permission change"
regex: '(?i)\bchmod\s+777\b'
- id: destructive-commands-2
severity: HIGH
pattern: destructive-commands
message: "Sudo escalation"
regex: '(?i)\bsudo\s+'
- id: destructive-commands-3
severity: HIGH
pattern: destructive-commands
message: "Disk overwrite command"
regex: '(?i)\bdd\s+if='
- id: destructive-commands-4
severity: HIGH
pattern: destructive-commands
message: "Filesystem format command"
regex: '(?i)\bmkfs\.'
# ── HIGH: dynamic code execution (JS/Python eval/exec) ──
- id: dynamic-code-exec-0
severity: HIGH
pattern: dynamic-code-exec
message: "Dynamic code execution (eval/exec) detected"
regex: '(?i)\b(eval|exec)\s*\('
exclude: '(?i)\b(eval|exec)\w+\s*\('
- id: dynamic-code-exec-1
severity: HIGH
pattern: dynamic-code-exec
message: "Dynamic Function constructor detected"
regex: '(?i)\bnew\s+Function\s*\('
# ── HIGH: Python shell execution ──
- id: shell-execution-0
severity: HIGH
pattern: shell-execution
message: "Python shell execution detected"
regex: '(?i)\b(os\.system|subprocess\.(run|call|Popen|check_output|check_call))\s*\('
# ── HIGH: hidden injection in HTML comments ──
- id: hidden-comment-injection-0
severity: HIGH
pattern: hidden-comment-injection
message: "Prompt injection hidden in HTML comment"
regex: '(?i)<!--.*?(ignore\s+(all\s+)?previous|disregard|you\s+are\s+now|forget\s+everything|override\s+safety|SYSTEM:)'
# ── HIGH: obfuscation ──
- id: obfuscation-0
severity: HIGH
pattern: obfuscation
message: "Base64 decode pipe may hide malicious content"
regex: '(?i)(base64\s+(-d|--decode)|\bbase64\b.*\|\s*(sh|bash|zsh|eval))'
# ── MEDIUM: escape sequence obfuscation ──
- id: escape-obfuscation-0
severity: MEDIUM
pattern: escape-obfuscation
message: "Consecutive escape sequences may hide content"
regex: '(\\x[0-9a-fA-F]{2}){3,}|(\\u[0-9a-fA-F]{4}){3,}'
# ── MEDIUM: suspicious URL usage ──
- id: suspicious-fetch-0
severity: MEDIUM
pattern: suspicious-fetch
message: "URL used in command context"
regex: '(?i)(curl|wget|invoke-webrequest|iwr)\s+[''"]?https?://'
exclude: '(?i)https?://(localhost|127\.0\.0\.1)'
# ── LOW: insecure transport hints ──
- id: insecure-http-0
severity: LOW
pattern: insecure-http
message: "Non-HTTPS URL in command context"
regex: '(?i)(curl|wget|invoke-webrequest|iwr)\s+http://'
exclude: '(?i)http://(localhost|127\.0\.0\.1|0\.0\.0\.0)'
# ── LOW: external links in markdown ──
- id: external-link-0
severity: LOW
pattern: external-link
message: "External URL in markdown link"
regex: '\]\(https?://[^)]+\)'
exclude: '(?i)\]\(https?://(localhost|127\.0\.0\.1|0\.0\.0\.0|learn\.microsoft\.com|developers\.figma\.com|developer\.mozilla\.org|lawsofux\.com|help\.figma\.com|developer\.apple\.com|help\.obsidian\.md|azure\.microsoft\.com|web\.dev|www\.microsoft\.com|opentelemetry\.io|nextjs\.org|motion\.dev|www\.radix-ui\.com|ui\.shadcn\.com|tailwindcss\.com|caniuse\.com)([:/)]|\?)'
# ── HIGH: remote code execution (pipe to interpreter) ──
- id: fetch-with-pipe-0
severity: HIGH
pattern: fetch-with-pipe
message: "Remote code execution: curl output piped to interpreter"
regex: '(?i)\bcurl\b[^|]*\|\s*(sudo\s+)?(ba)?sh\b'
- id: fetch-with-pipe-1
severity: HIGH
pattern: fetch-with-pipe
message: "Remote code execution: wget output piped to interpreter"
regex: '(?i)\bwget\b[^|]*\|\s*(sudo\s+)?(ba)?sh\b'
- id: fetch-with-pipe-2
severity: HIGH
pattern: fetch-with-pipe
message: "Remote code execution: fetched content piped to script interpreter"
regex: '(?i)\b(curl|wget)\b[^|]*\|\s*(sudo\s+)?(python[23]?|node|ruby|perl|zsh|fish)\b'
# ── MEDIUM: URL with raw IP address ──
- id: ip-address-url-0
severity: MEDIUM
pattern: ip-address-url
message: "URL with raw IP address — may bypass DNS-based security controls"
regex: 'https?://\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}([:/\s)]|$)'
exclude: '(?i)https?://(127\.0\.0\.1|0\.0\.0\.0|10\.\d{1,3}\.\d{1,3}\.\d{1,3}|172\.(1[6-9]|2\d|3[01])\.\d{1,3}\.\d{1,3}|192\.168\.\d{1,3}\.\d{1,3})([:/\s)]|$)'
# ── MEDIUM: data URI with embedded executable content ──
- id: data-uri-0
severity: MEDIUM
pattern: data-uri
message: "data: URI may embed executable or obfuscated content"
regex: '(?i)\]\(data:[^)]+\)'
# ── HIGH: hardcoded secrets (inline API keys, tokens, passwords) ──
- id: hardcoded-secret-0
severity: HIGH
pattern: hardcoded-secret
message: "Google API key detected"
regex: 'AIza[0-9A-Za-z\-_]{35}'
- id: hardcoded-secret-1
severity: HIGH
pattern: hardcoded-secret
message: "AWS access key detected"
regex: 'AKIA[0-9A-Z]{16}'
- id: hardcoded-secret-2
severity: HIGH
pattern: hardcoded-secret
message: "GitHub personal access token (classic) detected"
regex: 'gh[ps]_[A-Za-z0-9_]{36,}'
- id: hardcoded-secret-3
severity: HIGH
pattern: hardcoded-secret
message: "GitHub fine-grained personal access token detected"
regex: 'github_pat_[A-Za-z0-9_]{22,}'
- id: hardcoded-secret-4
severity: HIGH
pattern: hardcoded-secret
message: "Slack token detected"
regex: 'xox[bporas]-[A-Za-z0-9\-]{10,}'
- id: hardcoded-secret-5
severity: HIGH
pattern: hardcoded-secret
message: "OpenAI API key detected"
regex: 'sk-[A-Za-z0-9]{20,}T3BlbkFJ[A-Za-z0-9]{20,}'
- id: hardcoded-secret-6
severity: HIGH
pattern: hardcoded-secret
message: "Anthropic API key detected"
regex: 'sk-ant-[A-Za-z0-9\-_]{80,}'
- id: hardcoded-secret-7
severity: HIGH
pattern: hardcoded-secret
message: "Stripe secret or restricted key detected"
regex: '[sr]k_(live|test)_[A-Za-z0-9]{20,}'
- id: hardcoded-secret-8
severity: HIGH
pattern: hardcoded-secret
message: "Private key block detected"
regex: '-----BEGIN (RSA|DSA|EC|OPENSSH) PRIVATE KEY-----'
- id: hardcoded-secret-9
severity: HIGH
pattern: hardcoded-secret
message: "Hardcoded secret assignment detected"
regex: '(?i)(api[_-]?key|secret[_-]?key|access[_-]?token|auth[_-]?token|password)\s*[:=]\s*["\x27][A-Za-z0-9+/\-_]{20,}["\x27]'
# ── INFO: risky shell chaining patterns (advisory) ──
- id: shell-chain-0
severity: INFO
pattern: shell-chain
message: "Command chaining detected; verify intent"
regex: '(?m)(\|\||&&|;)\s*(rm|chmod|curl|wget|bash|sh)\b'
# Narrow imperative prose patterns; anchor at the start to avoid prohibitions.
- id: anti-refusal-0
severity: HIGH
pattern: anti-refusal
message: "Directive to comply despite safety policies"
confidence: 0.85
regex: '(?i)^\s*(?:(?:[-*+>]|\d+[.)])\s*)?(?:\*\*)?(?:never\s+refuse\s+any\s+request|always\s+(?:comply(?:\s+with\s+(?:any|all|every)\s+(?:user\s+)?requests?)?|answer(?:\s+(?:(?:any|all|every)\s+)?(?:user\s+)?questions?)?))(?:\*\*)?\s*[,]?\s*(?:regardless\s+of|despite|even\s+if)\s+(?:(?:any|all|the|your|its|their)\s+)?(?:safety|security|refusal)\s+(?:policies|policy|rules|restrictions|checks|guidelines|protocols|requirements)\b'
- id: prompt-leakage-0
severity: HIGH
pattern: prompt-leakage
message: "Directive to disclose hidden system instructions"
confidence: 0.85
regex: '(?i)^\s*(?:(?:[-*+>]|\d+[.)])\s*)?(?:\*\*)?(?:reveal|print|disclose|output|show)\s+(?:(?:your|the|hidden|original)\s+){1,3}(?:system\s+prompt|system\s+instructions)\b(?:\s+to\s+(?:the\s+)?user\b|\s*[.!]|\s*$)'
- id: memory-poisoning-0
severity: HIGH
pattern: memory-poisoning
message: "Directive to persist a safety override in memory"
confidence: 0.85
regex: '(?i)^\s*(?:(?:[-*+>]|\d+[.)])\s*)?(?:\*\*)?(?:remember\s+permanently|save\s+to\s+(?:long[- ]term\s+)?memory)\b.{0,100}\b(?:safety|security)\s+(?:checks|rules|policies|restrictions)\s+(?:are(?:\s+to\s+be)?|is(?:\s+to\s+be)?|should\s+be|must\s+be|gets)\s+(?:disabled|ignored|bypassed)\b'
- id: agent-snooping-0
severity: HIGH
pattern: agent-snooping
message: "Directive to disclose or access agent configuration or another skill without permission"
confidence: 0.85
regex: '(?i)^\s*(?:(?:[-*+>]|\d+[.)])\s*)?(?:\*\*)?(?:read|open|copy)\s+(?:~/(?:\.claude|\.codex|\.gemini)/\S+|mcp\.json|another\s+skill''s\s+SKILL\.md)\s+(?:and\s+(?:then\s+)?(?:send|share|exfiltrate)|without\s+(?:permission|consent))\b'
- id: persistence-0
severity: HIGH
pattern: persistence
message: "Directive to establish persistent execution"
confidence: 0.85
regex: '(?i)^\s*(?:(?:[-*+>]|\d+[.)])\s*)?(?:\*\*)?(?:install\s+a\s+cron\s+job\b|run\s+(?:crontab\s+-e|launchctl\s+load|systemctl\s+enable)\b|append\s+.{1,100}\s+to\s+~/\.(?:bashrc|zshrc|profile)\b).{0,100}\b(?:run|install|execute|launch|load)\s+(?:the\s+)?(?:payload\b(?:\s+(?:every|at|on|from|with)\b|\s*[.,;!]|\s*$)|(?:malware|backdoor)\b)|^\s*(?:(?:[-*+>]|\d+[.)])\s*)?(?:\*\*)?append\s+.{0,80}\b(?:payload|malware|backdoor)\b.{0,80}\s+to\s+~/\.(?:bashrc|zshrc|profile)\b'
# Sensitive destinations are evidence, not a runtime SSRF verdict. Metadata
# hosts deliberately remain visible in every admitted text format.
- id: ssrf-metadata-0
severity: CRITICAL
pattern: ssrf-metadata
message: "Cloud metadata host reference — review request context and possible credential exposure"
regex: '(?i)(?:^|[^a-z0-9_.])(?:169\.254\.(?:169\.254|170\.2)|100\.100\.100\.200|metadata\.google\.internal)(?:[^a-z0-9_.]|\.(?:[^a-z0-9_.]|$)|$)|(?:^|[^a-z0-9_.:]|[^0-9a-f:]:)fd00:ec2::254(?:[^a-z0-9_.:]|:(?:[^0-9a-f:]|$)|$)'
- id: ssrf-internal-0
severity: MEDIUM
pattern: ssrf-internal
message: "Request to a loopback, private, link-local or internal destination (static evidence)"
files: ['*.py', '*.js', '*.ts', '*.rb', '*.php', '*.php3', '*.php4', '*.php5', '*.phtml', '*.sh', '*.bash', '*.zsh', '*.fish']
regex: '(?:\b(?:requests|httpx|axios)\.(?:get|post|put|patch|delete|head|options)|\b(?:fetch|urlopen|file_get_contents|fopen)|\burllib\.request\.urlopen|\bNet::HTTP\.(?:get|post)|\bURI\.open)\s*\(\s*(?:URI\s*\(\s*)?["''](https?://[^"''\s]+)|\b(?:curl|wget)\b[^\n;|]*?["'']?(https?://[^"''\s;|)]+)'
- id: ssrf-dynamic-0
severity: MEDIUM
confidence: 0.6
pattern: ssrf-dynamic
message: "Dynamic request target — destination or validation unknown; review source and validation"
files: ['*.py', '*.js', '*.ts', '*.rb', '*.php', '*.php3', '*.php4', '*.php5', '*.phtml']
regex: '(?:\b(?:requests|httpx|axios)\.(?:get|post|put|patch|delete|head|options)|\b(?:fetch|urlopen|file_get_contents|fopen)|\burllib\.request\.urlopen|\bURI\.open)\s*\(\s*(?:(?:req|request)\.(?:query|body|params|args|GET)\b|\$_(?:GET|POST|REQUEST)\s*\[|params\s*\[|f["'']https?://\{|`https?://\$\{)'
- id: unsafe-deserialization-python
severity: HIGH
pattern: unsafe-deserialization
message: "Python pickle/marshal/dill deserialization can execute code with untrusted data"
files: ['*.py']
regex: '\b(?:pickle|marshal|dill)\.loads?\s*\('
- id: unsafe-deserialization-yaml
severity: HIGH
pattern: unsafe-deserialization
message: "Unsafe YAML loader or load call without an explicit SafeLoader/CSafeLoader"
files: ['*.py']
regex: '\byaml\.(?:load|load_all|unsafe_load|unsafe_load_all|full_load|full_load_all)\s*\('
- id: unsafe-deserialization-torch
severity: HIGH
pattern: unsafe-deserialization
message: "torch.load without explicit weights_only=True in this call"
files: ['*.py']
regex: '\btorch\.load\s*\('
- id: unsafe-deserialization-php
severity: HIGH
pattern: unsafe-deserialization
message: "PHP unserialize can instantiate objects from untrusted data"
files: ['*.php', '*.php3', '*.php4', '*.php5', '*.phtml']
regex: '(?i)\bunserialize\s*\('
- id: unsafe-deserialization-ruby
severity: HIGH
pattern: unsafe-deserialization
message: "Ruby Marshal.load/restore can execute code with untrusted data"
files: ['*.rb']
regex: '\bMarshal\.(?:load|restore)\s*\('
- id: unsafe-deserialization-js
severity: HIGH
pattern: unsafe-deserialization
message: "node-serialize dependency can deserialize executable values; review usage"
files: ['*.js', '*.ts']
regex: '(?:\brequire\s*\(\s*["'']node-serialize["'']\s*\)|\bfrom\s+["'']node-serialize["''])'
# File-level detector: thresholds and Markdown exemptions live in Go.
- id: whitespace-padding
severity: MEDIUM
pattern: whitespace-padding
message: "Padding may push instructions out of view; review context"