deepwork-writing-plans

Writes outcome-oriented implementation plans from approved specs before code changes begin.

Updated Jun 18, 2026
One-click install
npx skills add https://github.com/hugefiver/ocmm --skill deepwork-writing-plans-hugefiver
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deepwork-writing-plans
Source: https://github.com/hugefiver/ocmm/tree/main/plugins/deepwork/skills/deepwork-writing-plans
Command: npx skills add https://github.com/hugefiver/ocmm --skill deepwork-writing-plans-hugefiver

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Multi-step engineering work often starts coding before goals, dependencies, risks, and acceptance evidence are clear, leading to rework and missed requirements. This Skill turns an approved spec or requirements into a durable, file-backed implementation plan that a capable engineer can execute without rediscovering context. ## Core Features & Use Cases - Outcome-Oriented Plan Structure: Captures the approved goal, ideal end state, dependencies, interfaces, risks, assumptions, and acceptance evidence without prescribing rigid steps or commits. - Wave and Task Decomposition: Organizes work into dependency-ordered waves with observable end states and useful verification evidence for each. - Optional Evidence-Driven Review: Includes a plan-document reviewer prompt template for independent scrutiny when cross-module, security, or migration risk justifies it. - Use Case: Given an approved design for a multi-module migration, produce a plan saved to docs/superpowers/plans/ that defines waves, interfaces, risks, and acceptance evidence, then hand it off to executors or subagents. ## Quick Start Ask the agent to write an implementation plan for your approved spec before touching any code.

Frequently Asked Questions about deepwork-writing-plans

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

FAQPage Schema
How do I write an implementation plan before coding?▼

Read the approved design and repository evidence, then document the goal, ideal end state, dependencies, interfaces, risks, and acceptance evidence. Organize work into dependency-ordered waves and save the plan to docs/superpowers/plans/ as a dated Markdown file.

When should I write a file-backed plan instead of a todo list?▼

Use a durable plan when discovery shows unclear boundaries, cross-module coordination, novel behavior, migration or security risk, or enough work that executors need a shared artifact. For clear bounded changes, a concise in-chat plan or todo list is sufficient.

Does the plan require fixed steps, commits, or mandatory reviews?▼

No. Plans are outcome-oriented and scale with uncertainty and risk. Commits require explicit authorization, and plan review is optional, requested only for consequential risks, novel interfaces, or unresolved cross-module interactions.

Can executors deviate from the written plan?▼

Executors may choose evidence-based equivalents for incidental details, but must not change the approved goal, constraints, permissions, or acceptance criteria. Significant deviations and assumptions are recorded with reasons and the cost if wrong.

How does plan handoff to subagents work?▼

Hand off the approved outcome, plan text, dependencies, evidence, constraints, and recorded rulings. Dispatch uses the current callable schema, with a canonical GOAL, STOP WHEN, and EVIDENCE envelope for generic dispatch routes.