greploop
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it automatically processes external text that could be influenced by untrusted parties.
- Ingestion points: It fetches PR summaries, general comments, and inline discussion threads from GitHub, GitLab, and Perforce/Helix Swarm via CLI utilities and APIs (documented in
SKILL.md,references/gitlab-api.md, andreferences/graphql-queries.md). - Boundary markers: No delimiters or isolation strategies are present to distinguish between systemic automation commands and arbitrary text inside user or bot comments.
- Capability inventory: The skill possesses write access to code repositories and can modify files, execute commits, and push changes (
git push,p4 shelve) automatically for up to 5 iterations. - Sanitization: The retrieved text strings from reviews are parsed directly for status codes and code suggestions without screening or structural escaping.
Audit Metadata