code-comments
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and modifies code comments, which are untrusted external data sources that can be used to influence agent behavior.
- Ingestion points: The skill ingests comments from file diffs and user-provided source files as detailed in the 'Cleaning up or reviewing a change' section.
- Boundary markers: The instructions for calling subagents to review comments lack explicit delimiters or instructions to ignore embedded instructions in the processed data.
- Capability inventory: The agent is authorized to perform filesystem modifications (deleting lines, renaming code symbols, extracting variables) and can delegate tasks to subagents.
- Sanitization: There are no instructions to sanitize, escape, or filter comment content before it is processed by the AI.
- Note: The skill provides a natural mitigation by instructing the agent to treat common injection keywords like 'IMPORTANT' or 'do not remove' as claims to be verified rather than commands to be followed.
Audit Metadata