The Agent Skills Directory

[COMMAND_EXECUTION]: Employs standard git commands (diff, status, rev-parse, log, worktree, merge) to establish review context, manage isolation for fixes, and handle rollbacks.
[COMMAND_EXECUTION]: Executes local test suites (e.g., pytest, jest, cargo test, go test) during the verification phase to ensure code integrity and prevent regressions.
[SAFE]: Implements a multi-agent isolation model where initial audit agents are restricted to read-only tools (Read, Glob, Grep), preventing accidental or malicious changes during the discovery phase.
[SAFE]: Enforces a strict triage and confirmation flow; critical issues (P0) specifically including security vulnerabilities and exposed secrets require explicit user approval before any automated fix is attempted.
[SAFE]: Provides a robust 'Fix Loop' mechanism that uses isolated git worktrees for each modification, allowing for safe diff reviews and merges while maintaining the integrity of the main working branch.
[SAFE]: Includes a post-fix verification step that utilizes either existing project test runners or a dedicated read-only verification agent to check for syntax errors or broken dependencies.

deep-review