review-resolution

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from review comments, issues, and PRs. It maintains a strong security boundary by treating all external input as untrusted evidence and requiring independent technical validation before any code modification.
  • Ingestion points: SKILL.md Section 0 describes reading review sources via GitHub CLI.
  • Boundary markers: The skill includes a dedicated "Security Model" section and an invariant ("NEVER IMPLEMENT... BEFORE VERIFYING") to prevent the execution of malicious instructions embedded in review feedback.
  • Capability inventory: The skill performs read-only Git operations, read-only GitHub API calls, and localized file writes for validated fixes.
  • Sanitization: The workflow acts as a comprehensive validation and classification layer for external claims.
  • [COMMAND_EXECUTION]: The skill utilizes several standard command-line tools for environment inspection.
  • Git: Uses git status, git rev-parse, git branch, git diff, and git ls-files for state verification.
  • GitHub CLI: Uses gh api and gh pr view for fetching review content.
  • The instructions explicitly restrict these to read-only use and forbid writing to remote repositories or making additional network calls.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 04:21 AM