resolve-pr-parallel

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes git and the GitHub CLI (gh) to manage repository state, commit code changes, and resolve PR threads. These operations are essential to its function and restricted to the repository context.
  • [EXTERNAL_DOWNLOADS]: The skill references and uses scripts from the EveryInc/compound-engineering-plugin repository. This is an upstream vendor-aligned source used for the skill's logic.
  • [PROMPT_INJECTION]: The file references/evals.yaml includes test prompts with injection patterns like 'Ignore previous instructions'. These are utilized for security testing and are not executable vulnerabilities within the skill's logic.
  • [PROMPT_INJECTION]: The skill processes untrusted PR comments from GitHub, presenting a surface for indirect prompt injection.
  • Ingestion points: scripts/get-pr-comments fetches comment text from the GitHub API.
  • Boundary markers: No specific delimiters are used to wrap comment content during planning.
  • Capability inventory: File system writes, git commit, git push, and GitHub API mutations.
  • Sanitization: Relies on agent-level task classification and human-in-the-loop verification steps described in the workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 01:13 PM
Security Audit — agent-trust-hub — resolve-pr-parallel