github-commit-pr

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses standard, well-documented shell commands for Git (git add, git commit, git push, git pull) and GitHub CLI (gh pr create, gh run watch, gh pr merge). These are appropriate for the skill's stated purpose of managing the GitHub development lifecycle.
  • [DATA_EXPOSURE]: The skill actively prevents data exposure by implementing a mandatory sensitive file scan (references/sensitive-files.md) before staging changes. It explicitly instructs the agent to search for and warn the user about .env files, private keys, and credential stores.
  • [PRIVILEGE_ESCALATION]: The skill follows the principle of least privilege by explicitly prohibiting the use of gh pr merge --admin to bypass branch protections unless the user specifically requests it. It also prohibits skipping quality gates (e.g., --no-verify).
  • [REMOTE_CODE_EXECUTION]: While the skill mentions spawning subagents via the Task tool for heavy diff analysis, it specifies these must be read-only and that their output should be verified by the main agent. No remote script execution (e.g., curl | bash) is present.
  • [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the surface for indirect injection when processing CI logs or large diffs. It mitigates this by instructing the agent to treat subagent output as untrusted and to verify results against the actual source before acting.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 03:54 PM
Security Audit — agent-trust-hub — github-commit-pr