enact

Orchestrates multi-agent execution of design documents through coordinated subagent pipelines.

1|Updated Dec 18, 2009
One-click install
npx skills add https://github.com/thurn/dotfiles --skill enact-thurn
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: enact
Source: https://github.com/thurn/dotfiles/tree/main/.llms/skills/enact
Command: npx skills add https://github.com/thurn/dotfiles --skill enact-thurn

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Large multi-step software projects built from a design document require coordinated planning, implementation, review, and QA across many tasks. Doing this in a single context leads to lost state, inconsistent quality, and no durable record of progress. Enact turns a design document into a fully executed project by coordinating specialized subagents through a file-based run directory. ## Core Features & Use Cases - Orchestration Planning: Writes a standalone ENACT_PLAN.md with scope assessment, agent roster, per-task review/QA policy, and an executable manual QA strategy before any work begins. - Two-Phase Execution: Runs project-level agents (surveyor, researchers, planner, task generator, QA scenario generator) then a serial task-level cycle of feature coder, parallel conformance/quality reviewers and QA runner, and combined fix rounds. - File-Based State Management: Tracks ground truth in STATE.md, bugs in bugs/, reviews in reviews/, and QA results in qa/results/ under /tmp/enact/<run-name>/ so runs can resume after interruption. - Use Case: Given a design doc for a new CLI tool, invoke /enact to have an orchestrator plan the work, dispatch coding subagents per task, run code review and manual QA in parallel after each commit, fix findings, and finish with an integration review and retrospective. ## Quick Start Ask the AI to run /enact on your design document to plan and execute the full project with coordinated subagents.

Frequently Asked Questions about enact

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I execute a design document with multiple AI subagents?▼

Invoke the enact skill with your design document. The orchestrator writes an ENACT_PLAN.md, runs project-level agents for research and planning, then dispatches feature coder, reviewer, and QA subagents per task until the project passes integration review.

How does multi-agent orchestration keep track of progress across tasks?▼

Enact stores ground truth in STATE.md inside a run directory at /tmp/enact/<run-name>/. The orchestrator updates it after every step, so an interrupted run can resume by re-reading STATE.md and the artifacts under the run directory.

Can code review and manual QA run in parallel for each task?▼

Yes. After the feature coder commits, the conformance reviewer, quality reviewer, and QA runner inspect the same HEAD in parallel. Findings are collected and addressed in one combined fix round before targeted rechecks.

What happens when QA finds bugs during an orchestrated run?▼

The QA runner files bug reports under bugs/ in the run directory. The orchestrator includes them in the next fix round and reruns only the failed scenarios, capping the QA loop at three rounds per task before escalating to the user.

When should I not use multi-agent orchestration for a coding task?▼

Skip it for small, isolated changes with familiar patterns. Enact's scaling guidelines route low-complexity work through minimal agents, and the skill itself is intended only for large multi-step projects explicitly requested by name.