review-spec
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data derived from GitHub pull requests, creating a surface for indirect prompt injection attacks.
- Ingestion points: The agent reads content from
pr_diff.txtandpr_description.txt(SKILL.md). - Boundary markers: There are no explicit delimiters or protective instructions (e.g., XML tags or "ignore embedded instructions" markers) defined to isolate the untrusted snapshot data.
- Capability inventory: The skill allows reading repository files, writing a
review.jsonartifact, and executing a local Python validation script. - Sanitization: No sanitization, escaping, or validation of the input text is specified before the agent processes the content.
- [COMMAND_EXECUTION]: The skill invokes a local Python script (
python3 .github/skills/review-pr/scripts/validate_review_json.py) to perform validation on the generated review output. While the script is part of the repository infrastructure, it is executed against the untrustedpr_diff.txtdata.
Audit Metadata