receiving-code-review
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data in the form of external code review feedback from GitHub comments and threads.
- Ingestion points: Review feedback is ingested through GitHub API calls and CLI views, as documented in
references/pr-feedback-repair.md. - Boundary markers: The skill does not use specific structural delimiters for external content but establishes strict operational boundaries by requiring verification against the codebase before any action is taken.
- Capability inventory: The agent is expected to execute shell commands (
git,gh) and perform file system writes to implement repairs. - Sanitization: The skill provides strong defensive instructions, mandating that the agent "verify before implementing" and maintain skepticism toward external feedback to ensure technical correctness and avoid social engineering.
- [COMMAND_EXECUTION]: The skill provides templates for interacting with local and remote repositories via the command line.
- Evidence:
references/pr-feedback-repair.mdcontains specific shell patterns forgit fetch,git rev-parse, and a complexgh api graphqlquery used to extract structured metadata and comment threads from GitHub.
Audit Metadata