canon
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data such as Terms of Service, Privacy Policies, Data Processing Agreements (DPA), and audit evidence.
- Ingestion points: Processes user-supplied text for various legal and regulatory reviews across all files in the
reference/directory. - Boundary markers: The skill establishes clear role boundaries and 'Ask First' and 'Never' guidelines in
SKILL.mdto prevent the agent from making legal determinations or certifying compliance autonomously. - Capability inventory: The skill generates remediation instructions via the '## LLM Fix Prompt' block (detailed in
reference/fix-prompt-generation.md) and routes specific implementation tasks to other agents like Builder or Sentinel. - Sanitization: The instructions require the agent to use version-pinned citations from authoritative sources and structured checklists to validate all findings.
Audit Metadata