code-review
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface
- Ingestion points: The skill ingests untrusted data from local files through the execution of
git statusandgit diffcommands as outlined in the core workflow inSKILL.mdand thereferences/review-checklist.mdfile. - Boundary markers: There are no explicit delimiters or specific instructions for the agent to ignore or isolate instructions that may be embedded within the content of the files being reviewed.
- Capability inventory: The skill is explicitly restricted to read-only operations. Command execution is limited to non-mutating Git queries (
status,diff,diff --cached, etc.), and all state-changing operations are strictly forbidden and routed to external delivery skills. - Sanitization: No evidence of sanitization or content filtering is present in the instructions regarding the handling of interpolated file data during the review process.
Audit Metadata