finishing-branch

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local development and build tools such as pnpm, turbo, and uv (running pytest) for mandatory code verification steps.
  • [COMMAND_EXECUTION]: It performs various Git operations including branch switching, merging, and worktree management, as well as GitHub CLI commands (gh pr create) for PR lifecycle management.
  • [COMMAND_EXECUTION]: A hardcoded Windows path (D:\Node JS Starter V1) is used in the worktree cleanup step; while this restricts portability, it does not present a security risk.
  • [EXTERNAL_DOWNLOADS]: The skill interacts with GitHub's official services via git push and the gh CLI for branch and PR management. These are well-known technology services and appropriate for the skill's intended development workflow.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from git logs and branch names to populate PR descriptions, representing a surface for indirect prompt injection.
  • Ingestion points: git log and git diff outputs in SKILL.md.
  • Boundary markers: None explicitly defined for the interpolation of commit messages into the PR body.
  • Capability inventory: Shell execution of gh, git, and pnpm across SKILL.md.
  • Sanitization: None observed; the skill relies on the agent's context processing for generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 05:31 PM
Security Audit — agent-trust-hub — finishing-branch