What problem does it solve? It prevents AI coding agents from shipping partial, unverified, or assumption-based work by enforcing an absolute-certainty gate, mandatory planning, test-driven development, and manual QA before any task is declared done. ## Core Features & Use Cases - Absolute-certainty gate: Blocks implementation until requirements are fully understood, the codebase is explored, and all ambiguity is resolved. - Mandatory plan-agent orchestration: Delegates non-trivial tasks to a planning agent and executes the returned wave order, delegating work to specialist agents (explore, librarian, oracle, hephaestus, designer). - TDD red-green-surface workflow: Requires a failing test first, the smallest passing change, then a real-surface artifact (bash, curl, browser) proving the feature works. - Manual QA mandate: Rejects "tests pass" or "lsp clean" as sufficient evidence and demands end-to-end verification with cleanup of spawned resources. - Use Case: When implementing a new API endpoint, the agent writes a failing test, implements the minimal fix, curls the endpoint to capture the real response, and tears down any test resources before reporting completion. ## Quick Start Activate ultrawork mode with /ultrawork and ask the agent to implement your feature with full test-driven verification.
