deepwork

Orchestrates multi-phase coding workflows with delegated specialists and mandatory review gates.

Updated Jun 16, 2022
One-click install
npx skills add https://github.com/SimonZimmer/dotfiles --skill deepwork-simonzimmer
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deepwork
Source: https://github.com/SimonZimmer/dotfiles/tree/main/.config/opencode/skills/deepwork
Command: npx skills add https://github.com/SimonZimmer/dotfiles --skill deepwork-simonzimmer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Large, high-risk coding efforts with dependent phases often fail due to lost context, skipped reviews, and uncoordinated parallel work. This Skill provides a structured orchestration workflow that keeps plans, research, validation results, and review decisions in a persistent progress file. ## Core Features & Use Cases - Phased Planning and Delegation: Breaks large work into coherent delivery phases, assigns specialist lanes, and tracks ownership in a persistent markdown state file under .slim/deepwork/. - Mandatory Review Gates: Requires an Oracle review after each phase with a bounded re-review budget, plus structure scans when module boundaries change. - Scheduler Discipline: Coordinates background jobs, waits for completion events, and prevents advancing phases while results are unreconciled. - Use Case: When migrating a codebase across architectural boundaries, use this Skill to plan phases, delegate implementation to specialists, gate each phase with reviews, and commit only at validated delivery boundaries. ## Quick Start Ask the AI to run the deepwork workflow for a large multi-phase refactoring task and have it create a progress file, plan phases, and gate each phase with reviews.

Frequently Asked Questions about deepwork

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

FAQPage Schema
How do I manage a large multi-phase coding project with an AI agent?▼

Use an orchestrator workflow that splits work into coherent delivery phases, delegates implementation to specialist agents, and records state in a persistent progress file. Each phase ends with validation and a review gate before the next phase begins.

When should I use the deepwork workflow?▼

Use deepwork only for large or high-risk work: multiple dependent phases, cross-cutting architectural change, or migrations that are unsafe to partially ship. Do not use it for trivial edits, quick docs changes, simple bug fixes, or routine bounded features.

How do review gates work in a multi-phase coding workflow?▼

Each planned phase ends with a mandatory Oracle review. Every gate allows one initial review plus at most two re-reviews, requested only when remediation materially changes the reviewed decision. Exhausted re-reviews escalate remaining risk to the user.

Where does deepwork store its progress and state files?▼

Deepwork stores progress files under `.slim/deepwork/` as task-specific markdown files. The directory is added to `.gitignore` to stay local while `.ignore` rules keep it readable by the agent. Code deliverables go to normal project paths.

What are the limitations of an orchestrator-based coding workflow?▼

The orchestrator acts as a scheduler rather than the default implementer, so it adds overhead unsuited to small tasks. It also depends on specialist agents and review gates being available, and cannot advance phases while background jobs remain unreconciled.