What problem does it solve? Running a batch of independent stories one at a time is slow, but naive parallel execution risks corrupted shared state, colliding ADR/BUG numbers, and half-flipped story statuses. This Skill fans out one scoped sub-agent per Ready story in parallel, then reconciles all results on the main thread in a single serialized pass, so a batch of independent stories clears in one chat without breaking the board's gate contract. ## Core Features & Use Cases - Safety gate before fan-out: Verifies every story is DoR-ready and runs a static batch-check proving the batch is file-disjoint; any conflict, unknown, or script error falls back to serial execute-batch. - Scoped parallel sub-agents: Each sub-agent owns exactly one story and its testplan, is forbidden from touching MONITOR.md/ACTIVE.md, and returns a structured handoff instead of writing shared state. - Single-writer reconciliation: The main thread assigns sequential ADR/BUG numbers at write time, runs the per-story DoD gate, applies atomic status flips, writes the board once, and commits per story. - Honest partial-failure handling: Failed or crashed sub-agents leave their stories recoverable as in-progress or blocked, while succeeded stories still ship, with a shipped-vs-not report. - Use Case: Given CHAT-02 in an execution strategy with four file-disjoint Ready stories, invoke the command to implement and test all four concurrently, then fold results onto the board in one pass. ## Quick Start Run /tandem:execute-batch-parallel with a chat id such as CHAT-02 to execute that chat's Ready stories concurrently and reconcile the board.
