codex-review-loop
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses well-known, legitimate developer tools such as
gitand the GitHub CLI (gh). All network operations are confined to interactions with GitHub APIs, which are trusted services for project management and code review. - [PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection as it processes untrusted data from external pull request reviews.
- Ingestion points: Pull request comments and review summaries are fetched from GitHub through the
ghtool and specific API endpoints (SKILL.md). - Boundary markers: There are no specific delimiters used to isolate external review text from the agent's instructions.
- Capability inventory: The skill allows the agent to modify local source files, execute arbitrary shell commands for testing (e.g., build/lint), and push changes to a remote Git repository.
- Sanitization: The risk is mitigated by explicit instructions to treat review remarks as reports to be verified rather than commands to be followed, as well as the requirement to confirm fixes via local project tests before pushing code.
Audit Metadata