review-loop
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted code from a git repository, creating a potential surface for indirect prompt injection attacks. 1. Ingestion points: Processes repository files, diffs, and project history (SKILL.md, references/worker-briefs.md). 2. Boundary markers: Uses 'worker briefs' to guide sub-agents but lacks explicit instructions to ignore instructions embedded in the data being reviewed. 3. Capability inventory: Coordinates sub-agents through invoke-harness, writes to the file system, and runs validation scripts (SKILL.md). 4. Sanitization: Lacks specific sanitization or filtering of the repository data before processing.
- [EXTERNAL_DOWNLOADS]: The instructions include a fallback to fetch watch-list definitions from the author's GitHub repository (https://raw.githubusercontent.com/christopher-kapic/skills/master/skills/engineering/review-loop/references/watch/.md) when local access fails.
- [COMMAND_EXECUTION]: The skill coordinates the execution of validation commands, build scripts, and tests through runners during the implementation and fix cycles (SKILL.md, references/run-artifacts.md).
Audit Metadata