Rule set 2026.07.1

Rule Catalog

Every TrueTag finding carries the identifier and version of the rule that produced it. This page lists the 84 statically defined rules and the dynamic rule families in the current release, so a verdict can always be traced to the exact checks that ran.

A scan reports only what these rules can observe within the disclosed scope. “No findings” means no configured rule matched — it is not a certification that an application is secure, accessible, or legally compliant, and it does not replace penetration testing, professional accessibility testing, or legal advice.

Published corpus metrics

Precision and recall from the adjudicated CI corpus (36 cases, generated 2026-07-26). Launch-blocking families must stay at ≥ 98% precision on this corpus. Regenerated by pnpm evaluate:rules.

Family	Cases	Precision	Recall	Promotion
Secrets	25	100.0%	100.0%	allowed
Dependencies	20	100.0%	100.0%	allowed
Configuration	20	100.0%	100.0%	allowed

Committed secrets

stable

Deterministic credential patterns with entropy and placeholder guards. Matched values are redacted before storage and are never sent to a model provider.

Rule ID	Severity	What it detects
repo_secrets.aws_access_key	critical	Committed AWS access key ID
repo_secrets.aws_secret_key	critical	Committed AWS secret access key
repo_secrets.gcp_service_account	critical	Committed Google Cloud service-account key
repo_secrets.gcp_api_key	high	Committed Google API key
repo_secrets.stripe_live_secret	critical	Committed Stripe live secret key
repo_secrets.stripe_test_secret	medium	Committed Stripe test secret key
repo_secrets.stripe_webhook_secret	high	Committed Stripe webhook signing secret
repo_secrets.github_token	critical	Committed GitHub token
repo_secrets.github_pat_fine	critical	Committed GitHub fine-grained PAT
repo_secrets.openai_key	high	Committed OpenAI API key
repo_secrets.anthropic_key	high	Committed Anthropic API key
repo_secrets.slack_token	high	Committed Slack token
repo_secrets.sendgrid_key	high	Committed SendGrid API key
repo_secrets.twilio_key	high	Committed Twilio API key
repo_secrets.supabase_service_role	critical	Committed Supabase service-role JWT
repo_secrets.private_key_block	critical	Committed Private key material
repo_secrets.database_url	critical	Committed Database connection string with password
repo_secrets.jwt_secret_assignment	high	Committed Hard-coded JWT/session secret
repo_secrets.generic_api_key	high	Committed API key or token
repo_secrets.password_assignment	high	Committed Hard-coded password
repo_secrets.smtp_credentials	high	Committed SMTP credentials
repo_secrets.vercel_token	high	Committed Vercel token
repo_secrets.firebase_fcm_key	high	Committed Firebase Cloud Messaging server key
repo_secrets.bearer_token	high	Committed Hard-coded bearer token
repo_secrets.committed_env_file	high	Environment file committed to the repository

Dependencies and SBOM

stable

Package inventory across npm and pip manifests, advisory matching against the GitHub Advisory Database and OSV, and CycloneDX SBOM generation.

Rule ID	Severity	What it detects
repo_dependencies.vulnerable_package	varies	Dependency version matched a published security advisory
repo_dependencies.unpinned_dependency	low	Dependency accepts any version
repo_dependencies.missing_lockfile	medium	No lockfile committed

Authentication and authorization

beta

Heuristic checks for unauthenticated admin or mutating routes, bypass flags, and weak credential handling.

Rule ID	Severity	What it detects
repo_auth.admin_route_no_auth	critical	Admin route appears to lack an authentication check
repo_auth.mutating_route_no_auth	high	State-changing route without a recognized authentication check
repo_auth.hardcoded_email_check	medium	Authorization decided by a hard-coded email comparison
repo_auth.auth_bypass_flag	high	Authentication bypass flag present in code
repo_auth.token_in_localstorage	medium	Auth token stored in localStorage
repo_auth.weak_password_hashing	high	Passwords hashed with a fast, unsalted algorithm

Configuration and backend services

stable

Deployment configuration, CORS, and Firebase/Supabase backend policy checks.

Rule ID	Severity	What it detects
repo_config.debug_enabled	high	Debug mode enabled in configuration
repo_config.cors_wildcard	high	CORS allows any origin (severity rises when credentials are also allowed)
repo_config.localhost_in_config	low	Localhost URL in production-oriented configuration
repo_config.committed_source_maps	low	Production source maps committed to the repository
repo_config.firebase_open_rules	critical	Firebase rules allow public read or write access
repo_config.supabase_service_role_public	critical	Supabase service-role key exposed with a public env prefix
repo_config.supabase_rls_disabled	high	Row-level security disabled on a table
repo_config.supabase_permissive_policy	high	Supabase policy uses USING (true)

Input and output handling

beta

Injection and unsafe-rendering checks across request handling and templating.

Rule ID	Severity	What it detects
repo_input.dangerous_inner_html	high	Unsanitized HTML rendered with dangerouslySetInnerHTML
repo_input.inner_html_assignment	medium	Dynamic value assigned to innerHTML
repo_input.sql_injection	high	SQL query built by string concatenation
repo_input.command_injection	high	Shell command built from interpolated input
repo_input.path_traversal	high	File path built from request input without containment check
repo_input.open_redirect	medium	Redirect target taken directly from request input
repo_input.unrestricted_upload	medium	File upload without visible size or type restrictions

Payments

beta

Stripe integration checks for webhook verification and trusted amount handling.

Rule ID	Severity	What it detects
repo_payments.webhook_no_signature	critical	Stripe webhook handler does not verify the signature
repo_payments.client_amount	high	Payment amount taken from the client request
repo_payments.secret_key_client	critical	Stripe secret key exposed to the frontend

AI system security

beta

Prompt-injection surfaces, unsafe model-output handling, agency limits, and AI system-graph checks aligned with the OWASP LLM Top 10.

Rule ID	Severity	What it detects
ai_security.prompt_injection_surface	high	User input concatenated directly into a model prompt
ai_security.unsafe_output_handling	high	Model output may be rendered as HTML or executed
ai_security.excessive_agency	high	AI agent can take high-impact actions without a human check
ai_security.secret_in_prompt	medium	A secret appears to be placed inside a model prompt
ai_graph.retrieval_tenant_scope_unverified	high	Retrieval tenant isolation is not evident
ai_graph.high_impact_tool_without_approval	high	High-impact AI tool lacks a recognized approval gate
ai_graph.unvalidated_output_sink	high	Model output reaches a privileged sink without recognized validation
ai_graph.provider_disclosure_mismatch	medium	Detected AI provider is missing from the disclosed provider list

artifact.tt-art-* — Artifact Guard findings re-emitted for instruction and automation files found in the repository

Artifact Guard (pre-install)

stable

Deterministic checks applied to pasted or acquired AI skills, agent instructions, manifests, and tool definitions before installation.

Rule ID	Severity	What it detects
TT-ART-001	critical	Instruction hierarchy override
TT-ART-002	critical	Secret or environment-variable harvesting
TT-ART-003	critical	Downloaded content executed directly
TT-ART-004	high	Unrestricted command execution
TT-ART-005	high	Install-time execution hook
TT-ART-006	high	Instruction to disclose protected context
TT-ART-007	high	Hidden instruction in HTML comment
TT-ART-008	high	Visually concealed content
TT-ART-009	medium	Encoded or obfuscated payload
TT-ART-010	critical	Destructive filesystem command
TT-ART-011	high	Invisible Unicode control characters

Live-site transport, headers, cookies, and legal pages

stable

DNS-pinned, GET/HEAD-only probes of the running site: TLS, security headers, cookie flags, mixed content, exposed paths, and required legal pages.

Rule ID	Severity	What it detects
live_site.no_https	high	Site served over plain HTTP
live_site.tls_invalid	high	TLS certificate invalid
live_site.tls_expiring	medium	TLS certificate close to expiry
live_site.header_frame	low	No clickjacking protection (frame-ancestors or X-Frame-Options)
live_site.cookie_flags	medium	Cookie set without recommended security flags
live_site.mixed_content	medium	HTTPS page loads active content over HTTP
live_site.exposed_sensitive_path	varies	A sensitive path responds publicly
live_site.header_csp	medium	Missing Content-Security-Policy header
live_site.header_hsts	medium	Missing Strict-Transport-Security header
live_site.header_xcto	low	Missing X-Content-Type-Options header
live_site.header_referrer	low	Missing Referrer-Policy header
live_site.missing_privacy_policy	medium	No Privacy Policy link found on the homepage
live_site.missing_terms	low	No Terms link found on the homepage

Accessibility (WCAG 2.2 AA automation)

axe-core automated checks executed against the rendered page. Automated coverage is partial by nature; manual review remains necessary.

accessibility.<axe-rule-id> — One finding per violated axe-core rule, carrying the upstream rule identifier and WCAG mapping

External engines

Pinned external scanners (Gitleaks, OSV-Scanner, Trivy, Semgrep CE, YARA, and ClamAV) whose results are normalized into the same evidence model. Engine versions are recorded on every scan.

TT-EXT-<ENGINE>-<rule> — One finding per upstream engine rule hit, prefixed by the engine name