plan-mode

mpuig's avatarfrom mpuig

Guide for creating structured plans with numbered steps and quality gates

0stars🔀1forks📁View on GitHub🕐Updated Jan 10, 2026

When & Why to Use This Skill

This Claude skill provides a rigorous framework for generating structured, actionable implementation plans for complex technical workflows. By enforcing a standardized format—comprising deep situational analysis, strategic approach selection, atomic step-by-step execution, and mandatory quality gates—it ensures that software development and automation tasks are planned with high precision, minimizing errors and improving project transparency.

Use Cases

  • Complex Feature Implementation: Breaking down large-scale coding requirements into 3-10 manageable, sequential steps that specify file modifications and tool dependencies.
  • API and Tool Integration: Mapping out the integration of third-party services by defining necessary wrappers, configuration updates, and mock-testing strategies.
  • System Refactoring and Migration: Creating a safe roadmap for code changes that includes mandatory validation gates like unit testing, type checking, and dry runs to prevent regressions.
  • Automated Workflow Orchestration: Structuring the logic for multi-step agentic processes, ensuring that each phase passes specific quality benchmarks before proceeding.
nameplan-mode
descriptionGuide for creating structured plans with numbered steps and quality gates

Plan Mode Skill

Use this skill when creating a plan for workflow implementation.

Plan Structure

Your plan MUST follow this structure:

1. Analysis

  • What needs to be done and why?
  • What is the current state vs. desired state?
  • What are the constraints or requirements?

2. Approach

  • High-level strategy for implementation
  • Key design decisions
  • Which patterns or libraries to use

3. Steps

  • Numbered, concrete steps (3-10 steps typically)
  • Each step should be specific and actionable
  • Include which files will be modified
  • Mention which tools/imports are needed
  • Consider dependencies between steps

4. Quality Gates

  • List gates that should pass after execution
  • Always include: validate, dry
  • Add optional gates if needed: pytest, ruff, typecheck

Example Plan

## Analysis
The workflow needs to fetch stock prices from Yahoo Finance API to provide
real-time data. Currently, there's no data fetching capability.

## Approach
We will create a new tool `yahoo_finance` in tools/ that wraps the yfinance
library, then integrate it into the workflow using a @step decorator.

## Steps
1. Create tools/yahoo_finance/tool.py with fetch_stock_price() function
2. Add yfinance dependency to tool's config.yaml
3. Update workflow run.py to import yahoo_finance
4. Add @step("fetch_prices") that calls yahoo_finance.fetch_stock_price()
5. Update dry_run.py with mock stock data for testing
6. Add test case in test.py for the fetch step

## Quality Gates
- validate: Structural validation must pass
- dry: Dry run with mocks must complete successfully
- pytest: All unit tests must pass

Tips

  • Keep steps atomic and sequential
  • Don't skip validation in your plan
  • Always consider the dry run path
  • Think about error cases and edge conditions
  • Reference existing code patterns when possible
plan-mode – AI Agent Skills | Claude Skills