writing-plans Skill for Claude: Turn a Spec Into a Task Plan

writing-plans makes your agent turn a spec into a detailed, test-first implementation plan, saved as a file for you to review before any code is written.

Installs: 290.7K; on skills.sh
Stars: 310.6K; on GitHub
Updated: Oct 2026; last repo commit

Use writing-plans when you have a spec or requirements for a multi-step task, before touching code. The agent maps the files, then writes tasks as small steps: write the failing test, run it, implement, run it again, commit. Plans are saved to docs/superpowers/plans/YYYY-MM-DD-<feature>.md unless you set a different place.

Each plan starts with a header (goal, architecture, tech stack, spec, constraints, review focus) and ends with a five-point self-review. The agent then asks you to review the plan and choose subagent-driven or native execution. Both routes use other Superpowers skills, so install the whole set.

Good for

Developers with a spec who want a reviewable plan before coding starts
Teams handing work to subagents that have not seen the codebase
Anyone who wants each task to end with a testable result and a commit

Install writing-plans

Claude Code

Copy this prompt.

Install the writing-plans skill in Claude Code. Run this command:

/plugin install superpowers@claude-plugins-official

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Install the plugin for the whole Superpowers set. This skill hands off to superpowers:subagent-driven-development or superpowers:executing-plans, and mentions superpowers:using-git-worktrees.

From the project README. Source.

What's in the skill

When it loads

Use when you have a spec or requirements for a multi-step task, before touching code

Sections in SKILL.md

Overview
Scope Check
File Structure
Task Right-Sizing
Step Granularity
Plan Document Header
Global Constraints
Review Focus
Task Structure
Task N: [Component Name]
What a Step Contains
Self-Review
Execution Handoff

Key facts

Author: obra (obra/superpowers repo owner)
License: MIT
Installs on skills.sh: 290,684 as of 2026-11-02
GitHub stars: 310,635 for the whole obra/superpowers repo
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk pass
What is inside: SKILL.md only; bundles no other files
Needs: The Superpowers skills it hands off to (subagent-driven-development or executing-plans); a spec to plan from
Part of plugin: superpowers
Last commit to the repo: 2026-10-29

What it includes

Splits multi-system specs into separate plans
Suggests one plan per independent subsystem.
Try: “Use writing-plans for the spec in docs/specs/export.md.”
Read · Scope check
Maps files and their responsibilities first
Follows the codebase's existing patterns and prefers small, focused files.
Read · File structure
Files, interfaces, and 5 test-first steps
Each task lists files to create or modify, what it consumes and produces, then failing test, run, implement, run and commit.
Write · Task and step format
Saved under docs/superpowers/plans/
Starts with a header that names the required sub-skill for execution.
Write · Plan document
Spec coverage, step scan, types, review focus, proportion
The agent checks its own plan against the spec and fixes issues inline.
Read · Self-review
Asks you to pick subagent-driven or native
Subagent-driven uses superpowers:subagent-driven-development. Native uses superpowers:executing-plans. It waits for your review first.
Read · Execution handoff

From the official docs

Questions

What does writing-plans do?

It writes an implementation plan from a spec: files, interfaces, tests and small steps, ending in commits. It saves the plan as a markdown file and asks you to review it before any code is written.

How do I install it?

Use the install box at the top of this page, which has tabs for Claude Code, the Claude app and Codex. The install box uses the Superpowers plugin.

Where does it save the plan?

In docs/superpowers/plans/ with a date and feature name, unless you tell it to use another location.

What comes after the plan?

You choose subagent-driven execution, where a fresh subagent does each task and gets reviewed, or native execution in one session. SKILL.md names a Superpowers skill for each.

Does it write code?

It writes the plan, not the implementation. SKILL.md has the agent ask you to review the plan before implementation starts. The plan itself can include test code and exact signatures.

Does it work without the rest of Superpowers?

The plan header names superpowers:subagent-driven-development or superpowers:executing-plans as the required sub-skill for carrying it out, so install the whole set.

Is it safe?

It bundles no scripts and writes one plan file, but plans it writes can contain commands for later steps. See the audit results under key facts.

Information

updated Nov 2, 2026

Repository: obra/superpowers
Folder: skills/writing-plans

obra/superpowers skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2

obra/superpowers README · github.com
writing-plans SKILL.md · github.com