no-comments
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill treats code comments and diffs as untrusted data and interprets specific textual instructions found within them, such as 'do not remove' or 'IMPORTANT' markers.
- Ingestion points: Codebase file paths and git diffs are read in SKILL.md (Step 1) and passed to the 'Comment Sicko' sub-task described in comment-sicko.md.
- Boundary markers: The sub-task is restricted to a read-only brief, and the main skill includes an audit layer in Step 2 to review and reject incorrect findings before modification.
- Capability inventory: The skill possesses the ability to modify source code, delete comments and symbols (Steps 2, 3, 4), and invoke analytical tools such as 'how', 'why', and 'architect' via tool calls.
- Sanitization: Step 2 implements a verification process to filter findings, although it specifically directs the removal of safety and correctness suppressions if their justification is not immediately obvious to the agent.
Audit Metadata