Run a full spec implementation cycle — plan, track, implement, review, test, verify, and confirm. Orchestrates all skills and subagents in sequence with a review-fix-reverify loop.

Convert a technical specification into a sequenced implementation plan with discrete, trackable tasks. Reads the spec and codebase to produce an ordered plan.

Review how a cycle ran and improve automation skills. Collects signals from the tracker and git history, diagnoses issues, proposes changes, and A/B tests them via skill evals. Only keeps changes that measurably improve or maintain quality.

Brainstorm, create, or extend the project roadmap with new milestones. Use when the roadmap is exhausted, needs new versions, or when ideating future direction.

Display the current status of a spec's implementation tracker — shows task progress, blockers, and verification state.

Initialize or update a task tracker for a spec's implementation plan. Extracts tasks from the plan and creates a YAML tracker file with status, proof, and verification fields.

Generate a technical specification from a roadmap milestone. Reads the roadmap, design doc, and existing specs to produce a detailed spec for the next milestone.

Review code changes for correctness, spec compliance, and quality. Runs in isolated context to provide unbiased review. Reports issues that feed into the review-fix-reverify loop.

Write tests for implemented features based on spec acceptance criteria. Uses bun:test. Produces test files, runs them, and flags untestable criteria for ui-verifier.

Verify UI behavior and visual correctness using Claude Preview or Playwright. Starts the dev server, interacts with the playground, and captures screenshots as proof.

Verify that an implemented task meets its spec acceptance criteria. Reads the spec, inspects code changes, runs tests, and produces a pass/fail verdict with proof. Writes results directly to the tracker YAML.

4 hooks across 3 events