agent-tdd

A Claude Code plugin for human-agent co-authored TDD. You spec the test cases out loud; an orchestrator agent writes red tests, implements them, opens PRs, and merges waves of work in parallel.

You only ever talk to one agent. It runs the show.

Prerequisites

• Claude Code installed and authenticated.
• tmux ≥ 3.0.
• gh (GitHub CLI), authenticated for the target repo.
• git ≥ 2.5 (worktree support).
• A repo whose CI is reachable via gh pr checks --watch.

Install

Inside Claude Code:

/plugin marketplace add hn12404988/emacs_setup
/plugin install agent-tdd@willie-plugins

Use

1. In any tmux window, launch Claude Code.
2. Invoke the workflow:

   /agent-tdd:atdd <describe new feature or bug to the agent>

3. The agent discusses the spec with you (Wave 0). When you're aligned, say "go".
4. From there, the agent is in autopilot. The window title updates with live status, and you'll only be pinged when human input is genuinely needed.

Safety notes

• Impl agents run with --permission-mode bypassPermissions for non-interactive autonomy. Use only in trusted local repos.
• The workflow uses GitHub API heavily. Watch your 5000/hr authenticated rate limit on long workflows.
• Each parallel agent uses its own git worktree. For very large repos, N parallel agents ≈ N× working-tree disk.
• If the orchestrator process dies mid-workflow, you'll need to re-launch it manually — there is no automatic crash recovery in v1.

Learn more

• Design rationale → WHITEPAPER.md
• Operational spec → skills/atdd/PROTOCOL.md
• Status, known risks, future work → ROADMAP.md