What problem does it solve? Coordinating multiple AI subagents on parallel coding tasks often leads to conflicting edits, unverified claims of completion, and messy Git history. This Skill gives a supervising agent a disciplined workflow for splitting work, auditing subagent output against real evidence, and closing the loop on integration, testing, and GitHub issue cleanup. ## Core Features & Use Cases - Supervision Modes: Choose between watch-only monitoring, read-only audit, audit-and-fix, or multi-agent comparison depending on the user's request and permissions. - Evidence-Based Auditing: Verify subagent claims against git status, diffs, command output, CI results, and screenshots, classifying findings as Gap, Bug, Verification miss, Scope drift, or No issue. - GitHub Issue Parallelism: Isolate each issue in its own worktree and branch, then merge, test, clean up worktrees, push, and close issues through a controlled end-to-end procedure. - Use Case: A user asks to process three GitHub issues in parallel without mixing work; the supervising agent spawns one subagent per issue worktree, audits each result, merges branches with preflight conflict checks, runs focused tests, and closes only the verified issues. ## Quick Start Ask the agent to split this task across subagents and supervise them through to verified completion.
