commit-security-scan
Pass
Audited by Gen Agent Trust Hub on Mar 22, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data such as code commits, pull requests, and diffs. This exposes the agent to indirect prompt injection where instructions embedded in the analyzed content could override agent behavior.
- Ingestion points: The skill reads files and gathers context from the repository as specified in SKILL.md (Step 1) and may use external search/fetch tools as referenced in research-requirements.md.
- Boundary markers: No delimiters or explicit instructions to ignore embedded commands are present in the skill instructions.
- Capability inventory: The skill utilizes Read, Write, Bash, and Grep tools (SKILL.md).
- Sanitization: No sanitization, filtering, or validation of the untrusted input content is documented.
Audit Metadata