gsd

SKILL.md

Get Shit Done (GSD) - Project Management System

A comprehensive project management system designed for solo developers working with Claude AI agents. GSD provides structured workflows for project initialization, planning, execution, verification, and debugging.

Overview

GSD is a modular agent-based system that transforms project ideas into shipped software through:

Deep questioning - Extract user vision and requirements

Domain research - Discover standard stacks and patterns

Roadmap creation - Break requirements into phases

Phase planning - Create executable plans with verification

Execution - Implement plans with atomic commits

Verification - Ensure goals are achieved, not just tasks done

Debugging - Systematic investigation of issues

Core Philosophy

Solo Developer + Claude Workflow - No teams, no stakeholders, no ceremonies

Plans are Prompts - PLAN.md files ARE the execution prompts, not documents

Goal-Backward Planning - Start from what must be TRUE, derive what to build

Atomic Commits - Each task commits independently for clean history

Quality Over Speed - Stop before context degrades (~50% usage)

Ship Fast - Plan → Execute → Ship → Learn → Repeat

When to Use GSD

Use GSD when you need to:

Initialize a new software project

Plan and execute development phases

Map an existing codebase

Debug issues systematically

Verify phase completion

Track project progress and state

Quick Start Commands

New Project

/gsd:new-project

Initializes a new project with questioning → research → requirements → roadmap flow.

Plan a Phase

/gsd:plan-phase [phase-number]

Creates detailed execution plans with research and verification.

Execute a Phase

/gsd:execute-phase [phase-number]

Executes all plans in a phase with parallel execution support.

Map Codebase

/gsd:map-codebase [optional-focus-area]

Analyzes existing codebase with parallel mapper agents.

Debug Issues

/gsd:debug [issue-description]

Systematic debugging using scientific method and hypothesis testing.

Verify Phase

/gsd:verify-work [phase-number]

Goal-backward verification of phase completion.

Check Progress

/gsd:progress

Shows current project position, completed phases, and next steps.

Agent Skills

GSD includes specialized agents for different tasks:

gsd-codebase-mapper - Explores and documents codebase structure

gsd-planner - Creates executable phase plans

gsd-executor - Executes plans with atomic commits

gsd-debugger - Investigates bugs systematically

gsd-verifier - Verifies goal achievement

gsd-research-synthesizer - Synthesizes research outputs

gsd-roadmapper - Creates project roadmaps

gsd-phase-researcher - Researches phase implementation

gsd-project-researcher - Researches domain ecosystem

gsd-integration-checker - Verifies integrations work

gsd-plan-checker - Validates plan quality

Command Skills

GSD provides commands for orchestrating the entire project lifecycle:

gsd:new-project - Initialize new project

gsd:map-codebase - Map existing codebase

gsd:plan-phase - Plan a phase

gsd:execute-phase - Execute a phase

gsd:verify-work - Verify phase completion

gsd:debug - Debug issues

gsd:discuss-phase - Gather phase context

gsd:research-phase - Research phase implementation

gsd:complete-milestone - Complete milestone

gsd:audit-milestone - Audit milestone

gsd:add-phase - Add new phase

gsd:insert-phase - Insert phase

gsd:remove-phase - Remove phase

gsd:add-todo - Add todo item

gsd:check-todos - Check todos

gsd:plan-milestone-gaps - Plan milestone gaps

gsd:pause-work - Pause work

gsd:resume-work - Resume work

gsd:update - Update project state

gsd:whats-new - Show what's new

Workflow Skills

Detailed workflow definitions for complex operations:

discovery-phase - Phase discovery workflow

execute-phase - Phase execution workflow

diagnose-issues - Parallel UAT diagnosis

map-codebase - Codebase mapping workflow

discuss-phase - Phase discussion workflow

verify-phase - Phase verification workflow

verify-work - Work verification workflow

transition - Phase transition workflow

resume-project - Project resumption workflow

Reference Skills

Reference documents for best practices and guidelines:

questioning - Deep questioning techniques

tdd - Test-driven development patterns

ui-brand - UI/UX guidelines

verification-patterns - Verification methodologies

git-integration - Git workflow patterns

checkpoints - Checkpoint handling

continuation-format - Continuation format specification

Project Structure

GSD creates a .planning/ directory with:

.planning/
├── PROJECT.md           # Project context and vision
├── config.json          # Workflow preferences
├── REQUIREMENTS.md      # Scoped requirements
├── ROADMAP.md          # Phase structure
├── STATE.md            # Project memory and state
├── research/            # Domain research outputs
├── phases/              # Phase-specific artifacts
│   ├── XX-name/
│   │   ├── XX-PLAN.md
│   │   ├── XX-SUMMARY.md
│   │   ├── XX-CONTEXT.md
│   │   ├── XX-RESEARCH.md
│   │   ├── XX-VERIFICATION.md
│   │   └── XX-UAT.md
└── codebase/            # Codebase analysis
    ├── STACK.md
    ├── ARCHITECTURE.md
    ├── STRUCTURE.md
    ├── CONVENTIONS.md
    ├── TESTING.md
    ├── INTEGRATIONS.md
    └── CONCERNS.md

Key Concepts

Goal-Backward Planning

Instead of asking "what should we build?", ask "what must be TRUE for the goal to be achieved?"

Forward: "Build authentication system" → task list
Goal-Backward: "Users can securely access accounts" → derive what must exist

Atomic Commits

Each task commits independently with descriptive messages:

feat(01-01): implement user login
fix(01-02): fix password validation
test(01-03): add login tests

Context Budgeting

Plans complete within ~50% context usage to maintain quality:

0-30%: PEAK quality

30-50%: GOOD quality

50-70%: DEGRADING quality

70%+: POOR quality (avoid)

Wave-Based Execution

Plans are grouped into waves for parallel execution:

Wave 1: Independent plans (no dependencies)

Wave 2: Plans depending only on Wave 1

Wave 3: Plans depending on Wave 2, etc.

Anti-Patterns to Avoid

Enterprise PM Theater - No RACI matrices, sprint ceremonies, stakeholder management

Horizontal Layers - Don't group by "all models, then all APIs" - group by features

Vague Success Criteria - "Authentication works" → "User can log in with email/password"

Time Estimates - Never estimate in hours/days/weeks

Task Completion ≠ Goal Achievement - Verify outcomes, not just task completion

Getting Help

Each agent, command, and workflow has its own SKILL.md with detailed instructions. Use:

@skills/gsd/agents/ for agent-specific help

@skills/gsd/commands/ for command-specific help

@skills/gsd/workflows/ for workflow-specific help

@skills/gsd/references/ for reference documentation

Version

GSD Version: 1.0.0
Last Updated: 2026-01-19