The Agent Skills Directory

[SAFE]: The skill implements a multi-perspective review process using specialist sub-agents (code-quality, architecture, security, test-coverage) with isolated contexts to ensure objective analysis.
[SAFE]: Includes a robust human-in-the-loop validation phase (Phase 3: Human UAT) that requires manual confirmation of deliverables, preventing automated progression through failures.
[SAFE]: The finishing phase uses standard, well-known development tools such as the GitHub CLI (gh) for PR creation and git for worktree management.
[SAFE]: Security is a primary focus of the skill, with a dedicated specialist agent designed to detect OWASP Top 10 vulnerabilities, injection flaws, and hardcoded secrets in the code being reviewed.
[SAFE]: Instructions correctly utilize boundary markers (delimiters) when passing untrusted data like git diffs to sub-agents, adhering to prompt engineering best practices for data isolation.

reviewing