What problem does it solve? Getting a single AI agent to review task planning documents can miss issues, but running multiple reviewers in parallel risks conflicting edits and duplicate comments in the same files. This Skill coordinates three independent reviewers without any of them touching the docs, then de-duplicates and merges their findings into one clean set of comment threads. ## Core Features & Use Cases - Parallel Multi-Agent Review: Launches Claude Code, Codex, and Grok as concurrent read-only reviewers over the planning docs in docs/tasks/, so wall-clock time equals the slowest single reviewer. - Deterministic De-duplication: The coordinator is the single writer; it pools structured JSON findings, merges duplicates across reviewers, keeps the highest priority, and tags each thread with reviewer attribution (e.g. [council: claude, codex]). - Markdown and HTML Comment Threads: Appends append-only comment threads in either markdown comment syntax or <pd-thread> elements, then hands off to resolve-comments and verifies author replies. - Use Case: Before executing task 042, run a council review so three independent agents verify acceptance criteria, file paths, and test coverage against the actual codebase, with consensus signaling which issues are real. ## Quick Start Ask the agent to run a council review on task 042 so Claude, Codex, and Grok review the planning docs in parallel and the merged comments get resolved.
