The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or unauthorized data access found. The skill operates within a defined scope for code analysis and follows security best practices.\n- [COMMAND_EXECUTION]: The skill utilizes git and gh (GitHub CLI) to inspect codebase state and pull request metadata. These operations are restricted by the allowed-tools policy in the frontmatter to specific binary patterns (git *, gh *), which are appropriate for its function.\n- [DATA_EXFILTRATION]: While the skill interacts with GitHub via the gh tool, no evidence of exfiltration to third-party domains or unauthorized access to sensitive files (such as .ssh/ or .env files) was detected. Interactions are limited to repository metadata.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to process untrusted code files. However, the risk is mitigated by explicit anti-hallucination and grounding guidelines.\n
Ingestion points: Reads code files and diffs via Read, Grep, Glob, and gh pr diff (SKILL.md).\n
Boundary markers: Explicit boundary markers for processed code are absent, though role instructions define a strict senior developer reviewer persona.\n
Capability inventory: The skill has access to Write, Edit, and limited Bash (git/gh) operations.\n
Sanitization: No explicit sanitization of input code content is documented.

forge-review