address-pr-feedback
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources that could contain malicious instructions designed to subvert the agent's logic.\n
- Ingestion points:
scripts/fetch-feedback.shretrieves pull request comment bodies and review content;scripts/fetch-checks.shretrieves GitHub Actions job logs.\n - Boundary markers: The documentation provides procedural boundaries (e.g., distinguishing bot vs. human feedback) but does not mandate technical delimiters to isolate untrusted content within the LLM's context.\n
- Capability inventory: The agent can modify repository code, execute tests/build commands, and perform GitHub actions such as replying to and resolving threads via
scripts/resolve-thread.sh.\n - Sanitization: The provided scripts do not perform sanitization or filtering of the fetched text before it is presented to the model.
Audit Metadata