The Agent Skills Directory

[SAFE]: The skill utilizes standard git commands including git diff, git log, and git rev-parse to retrieve repository information for code review purposes.
[SAFE]: Repository documentation such as CODING_STANDARDS.md and specification files are used as reference. No patterns of unauthorized data exfiltration or hardcoded credentials were detected.
[PROMPT_INJECTION]: The skill processes untrusted input from git diffs, commit logs, and specification documents, which could contain malicious instructions designed to bias the sub-agents' review. (1) Ingestion points: Git diff output, commit logs, and specification files. (2) Boundary markers: Not explicitly defined in sub-agent prompts. (3) Capability inventory: Spawning sub-agents via the Agent tool and reading local files. (4) Sanitization: Not performed on the ingested content. This risk is common in text analysis skills and is partially mitigated by the segmented review architecture.

review