to-raw-requirements
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from chat history and existing local files (docs/requirements/raw.md) to generate structured documentation. This presents a potential surface for indirect prompt injection if those sources contain malicious instructions.\n
- Ingestion points: User-provided context and the current raw.md file as specified in Step 1 of the workflow.\n
- Boundary markers: The workflow uses a manual extraction and verification process in Step 2 and 4 to isolate specific information types, acting as a logical filter before data is written.\n
- Capability inventory: The skill has capabilities to write to the local file system and request Git commits as described in Step 5 and 6.\n
- Sanitization: The skill enforces a strict Markdown template defined in references/requirements-structure.md and explicitly forbids the agent from making up information or links, requiring all data to be confirmed by the user.\n- [COMMAND_EXECUTION]: The skill contains instructions to perform repository management actions.\n
- Evidence: Step 6 in SKILL.md requires the agent to obtain explicit user consent before executing a git commit command, ensuring that repository changes are user-initiated and reviewed.
Audit Metadata