I got tired of coding agents editing before “done” was defined — so I built a plan-first harness

Description and use

Loomwright introduces a plan-first approach for agent-driven development, creating a brief, isolating changes in git worktrees, and parking PRs for review before merge, reducing conflicts and improving automation reliability.

Problem

Agents edit code without a shared definition of done, leading to overlapping changes, merge conflicts, and untrusted automation that may silently merge incorrect code.

Outcome

Loomwright reduces merge conflicts, enforces a review gate, and ensures code changes are only merged after explicit approval, improving reliability of agent-driven development.

Workflow

• 1. Run /setup to install Loomwright and view module status.
• 2. Execute /autonomous with a feature prompt; the system asks discovery questions and saves a brief in .supervisor/jobs/pending/.
• 3. Loomwright creates a git worktree, runs workers, and generates an AUTONOMOUS_RUN summary.
• 4. An open PR is created; it remains parked for review and does not merge automatically.
• 5. Optionally, use --auto-merge behind a trusted gate to merge after review.

Evidence

Launch Pad writes a brief before Supervisor touches the repo

Source

Workers run in git worktrees; overlapping files wait; merges stay ordered

Source