code-review

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a 'remember' command that persists external context (standards, documentation links, and notes) into ~/.config/agent-skills/code-review.yaml. This context is automatically loaded and factored into the agent's reasoning during subsequent reviews. If this context contains malicious instructions or biased architectural constraints from an untrusted source, it could influence the agent's review findings or verdict.
  • Ingestion points: User-provided text and URLs via the remember command in SKILL.md (Workflow Step 0).
  • Boundary markers: None identified for the loaded context; the agent is instructed to "keep it in mind throughout the review" and "actively check compliance."
  • Capability inventory: Executes shell commands (git, gh, glab) and calls external scripts (github.py, gitlab.py, gerrit.py) as seen in SKILL.md Steps 1-6.
  • Sanitization: No explicit sanitization or validation of the remembered context is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 11:06 AM
Security Audit — agent-trust-hub — code-review