Agent skill

plan-writing

Structured task planning with clear breakdowns, dependencies, and verification criteria. Use when implementing features, refactoring, or any multi-step work.

View SKILL.md on GitHub Repository
Stars 23,776
Forks 2,298

Install this agent skill to your Project

npx add-skill https://github.com/davila7/claude-code-templates/tree/main/cli-tool/components/skills/productivity/plan-writing

SKILL.md

Plan Writing

Source: obra/superpowers

Overview

This skill provides a framework for breaking down work into clear, actionable tasks with verification criteria.

Task Breakdown Principles

1. Small, Focused Tasks

  • Each task should take 2-5 minutes
  • One clear outcome per task
  • Independently verifiable

2. Clear Verification

  • How do you know it's done?
  • What can you check/test?
  • What's the expected output?

3. Logical Ordering

  • Dependencies identified
  • Parallel work where possible
  • Critical path highlighted
  • Phase X: Verification is always LAST

4. Dynamic Naming in Project Root

  • Plan files are saved as {task-slug}.md in the PROJECT ROOT
  • Name derived from task (e.g., "add auth" → auth-feature.md)
  • NEVER inside .claude/, docs/, or temp folders

Planning Principles (NOT Templates!)

🔴 NO fixed templates. Each plan is UNIQUE to the task.

Principle 1: Keep It SHORT

❌ Wrong ✅ Right
50 tasks with sub-sub-tasks 5-10 clear tasks max
Every micro-step listed Only actionable items
Verbose descriptions One-line per task

Rule: If plan is longer than 1 page, it's too long. Simplify.

Principle 2: Be SPECIFIC, Not Generic

❌ Wrong ✅ Right
"Set up project" "Run npx create-next-app"
"Add authentication" "Install next-auth, create /api/auth/[...nextauth].ts"
"Style the UI" "Add Tailwind classes to Header.tsx"

Rule: Each task should have a clear, verifiable outcome.

Principle 3: Dynamic Content Based on Project Type

For NEW PROJECT:

  • What tech stack? (decide first)
  • What's the MVP? (minimal features)
  • What's the file structure?

For FEATURE ADDITION:

  • Which files are affected?
  • What dependencies needed?
  • How to verify it works?

For BUG FIX:

  • What's the root cause?
  • What file/line to change?
  • How to test the fix?

Principle 4: Scripts Are Project-Specific

🔴 DO NOT copy-paste script commands. Choose based on project type.

Project Type Relevant Scripts
Frontend/React ux_audit.py, accessibility_checker.py
Backend/API api_validator.py, security_scan.py
Mobile mobile_audit.py
Database schema_validator.py
Full-stack Mix of above based on what you touched

Wrong: Adding all scripts to every plan Right: Only scripts relevant to THIS task

Principle 5: Verification is Simple

❌ Wrong ✅ Right
"Verify the component works correctly" "Run npm run dev, click button, see toast"
"Test the API" "curl localhost:3000/api/users returns 200"
"Check styles" "Open browser, verify dark mode toggle works"

Plan Structure (Flexible, Not Fixed!)

# [Task Name]

## Goal
One sentence: What are we building/fixing?

## Tasks
- [ ] Task 1: [Specific action] → Verify: [How to check]
- [ ] Task 2: [Specific action] → Verify: [How to check]
- [ ] Task 3: [Specific action] → Verify: [How to check]

## Done When
- [ ] [Main success criteria]

That's it. No phases, no sub-sections unless truly needed. Keep it minimal. Add complexity only when required.

Notes

[Any important considerations]


---

## Best Practices (Quick Reference)

1. **Start with goal** - What are we building/fixing?
2. **Max 10 tasks** - If more, break into multiple plans
3. **Each task verifiable** - Clear "done" criteria
4. **Project-specific** - No copy-paste templates
5. **Update as you go** - Mark `[x]` when complete

---

## When to Use

- New project from scratch
- Adding a feature
- Fixing a bug (if complex)
- Refactoring multiple files

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

davila7/claude-code-templates

verl-rl-training

Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.

23,776 2,298
Explore
davila7/claude-code-templates

openrlhf-training

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

23,776 2,298
Explore
davila7/claude-code-templates

gguf-quantization

GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.

23,776 2,298
Explore
davila7/claude-code-templates

Claude Code Guide

Master guide for using Claude Code effectively. Includes configuration templates, prompting strategies "Thinking" keywords, debugging techniques, and best practices for interacting with the agent.

23,776 2,298
Explore
davila7/claude-code-templates

qdrant-vector-search

High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.

23,776 2,298
Explore
davila7/claude-code-templates

behavioral-modes

AI operational modes (brainstorm, implement, debug, review, teach, ship, orchestrate). Use to adapt behavior based on task type.

23,776 2,298
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results