resolve-review-comments
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted feedback from external pull or merge requests which could contain malicious instructions designed to bypass agent constraints.
- Ingestion points: Fetches comprehensive review context including threads, comments, and summaries from GitHub and GitLab APIs.
- Boundary markers: Implements a mandatory triage gate and item-level ledger to classify every actionable request.
- Capability inventory: Capability to execute shell commands (git, gh, glab), modify source code, and push to remote repositories.
- Sanitization: Employs a semantic triage logic with explicit classification (Implement, Skip, Clarify) and requires user confirmation in guarded mode.
- [COMMAND_EXECUTION]: The skill relies on system-level CLI tools for repository operations and platform metadata management.
- Evidence: Executes the GitHub CLI (gh) and GitLab CLI (glab) for PR/MR interactions.
- Evidence: Uses standard git commands for branch management, commits, and pushing changes to source repositories.
- [EXTERNAL_DOWNLOADS]: The skill may prompt the user to install necessary external dependencies if they are not detected on the host system.
- Evidence: Identifies missing gh or glab CLIs and requests explicit authorization for installation.
Audit Metadata