Matt Pocock pipeline variant that dispatches implementation to Codex (via /codex:rescue + mp-implement-issue) and code-quality review to Codex (via /codex:adversarial-review), keeping only spec-compliance review as Claude Code subagents. Planning (Phase 0-2 + Phase 4) stays in this session. Use when the user wants a…
Use when the user provides any kind of feature spec (rough notes / system design / partial spec / canonical PRD) and wants it implemented end-to-end autonomously on the Kimi Code CLI (AgentSwarm-engined), or invokes /kimi-mp-implement.
Pick up a single AFK-labeled issue (from a PRD already decomposed by /to-tickets or /mp-implement) and implement it end-to-end using Matt Pocock's TDD + diagnose discipline. Use when user wants to drive one issue at a time, says "implement.
Use when the user provides any kind of feature spec (rough notes / system design / partial spec / canonical PRD) and wants it implemented end-to-end autonomously on Claude Code alone — no external execution engine — or invokes /mp-implement.
Use when receiving code review feedback, before implementing suggestions — verify each finding against the codebase, clarify unclear items first, act by severity, push back with evidence when the reviewer is wrong.
Execute a decomposed plan in-session — fresh subagent per task, fixed spec-then-quality review loops. /mp-implement's Phase 3 fallback engine when the Workflow tool is unavailable.
Use when starting feature work that needs isolation from the current workspace, or before executing implementation plans — ensures an isolated workspace exists via native worktree tools or git worktree fallback.
Run the verification and read its output before claiming work is complete, fixed, or passing — evidence first, claim second. Use before any completion claim, commit, PR, or harvest.