project-bootstrap

name	project-bootstrap
description	Bootstrap Claude Code setup for greenfield (new) or brownfield (existing) projects. Automatically detects project type, copies relevant skills from your skill library, and creates engineer agents with best practices skills. Use when starting work on a new project or wanting to upgrade an existing project's Claude Code setup with skills and agents.

Project Bootstrap

Automated setup skill for bootstrapping Claude Code in any project. Detects whether you're in a greenfield or brownfield project, identifies the tech stack, copies relevant skills from your skill library, and creates specialized engineer agents.

Configuration

This skill uses hardcoded paths for personal use:

• Skill Library: C:\Users\coper\Documents\GitHub\agent-skills\skills
• Unpack Destination: .claude/skills/ in the current project

Workflow

Phase 1: Project Detection

1. Check if .claude/ exists (indicates brownfield with some setup)
2. Check for common project files: package.json, pyproject.toml, go.mod, Cargo.toml, etc.
3. If no project files found → greenfield project
4. If project files found → brownfield project

Phase 2: Information Gathering

For Greenfield Projects:

1. Search for README.md, specs, or planning documents
2. If found, extract project intent from those files
3. If not found or unclear, use AskUserQuestion to interview user about:
   • What are you building?
   • Primary language/framework?
   • Backend, frontend, or full-stack?
   • Any specific libraries or tools needed?

For Brownfield Projects:

1. Spawn Explore sub-agent to analyze codebase (stack, structure, patterns)
2. Spawn Explore sub-agent to check existing .claude/ setup (agents, skills, commands)
3. Wait for both reports before proceeding

Phase 3: Skill Selection

1. Read references/skill-categories.md for available skills mapping
2. Match project stack to relevant skills using keyword matching
3. Always include best-practices skill if available
4. Present selections to user via AskUserQuestion for confirmation
5. User can add/remove skills from the selection

Phase 4: Skill Deployment

1. Create .claude/skills/ directory if needed
2. Copy selected .skill files from skill library
3. Unpack each skill (extract zip to folder, delete .skill file)
4. Verify unpacked structure (SKILL.md exists)

Phase 5: Agent Creation

Automatically create agents without further confirmation (per user preference).

CRITICAL: Agent Prompt Quality

Each agent must have a rich, comprehensive system prompt following the agent creation architect framework. DO NOT create basic or minimal prompts. Each agent prompt should include:

1. Expert Persona: A compelling identity with deep domain knowledge
2. Core Intent: Clear purpose and success criteria
3. Comprehensive Instructions: Methodologies, best practices, edge case handling
4. Decision Frameworks: How to approach choices and trade-offs
5. Quality Control: Self-verification steps and output expectations
6. Workflow Patterns: Step-by-step process for task execution

Agent Types to Create:

Engineer Agents (1-3 based on stack):

• Create stack-specific engineers (e.g., typescript-engineer, python-engineer)
• Include best-practices skill reference
• Include relevant library skills
• Use rich prompts from references/agent-templates.md

Code Quality Agents:

• Create code-quality agent for the primary language
• Include linting/formatting skills (biome, ruff, etc.)
• Include best-practices skill
• Define clear quality checklist and verification steps

Library-Specific Agents (if applicable):

• Create agents for major frameworks (react, angular, express, etc.)
• Only if corresponding skill was copied
• Keep minimal (don't create agent for every skill)
• Include framework-specific patterns and best practices

See references/agent-templates.md for comprehensive agent creation patterns with rich prompts.

Phase 6: Validation & Summary

1. Run scripts/validate_setup.py to check all skills and agents
2. Report any issues found
3. Print summary of what was set up:
   • Skills copied and unpacked
   • Agents created
   • Any issues or warnings

Decision Framework

When to Ask vs. Automate

Situation	Action
Greenfield with no specs	Ask about project intent
Skill selection	Always confirm with user
Agent creation	Fully automatic
Unclear stack	Ask for clarification
Multiple valid approaches	Ask for preference

Stack Detection Keywords

Keyword	Skills to Suggest
typescript, javascript	typescript-best-practices, biome, vite
react	react-19, react-ecosystem skills
python	python-best-practices, ruff-dev, pytest
go	go-best-practices
rust	rust-dev
angular	angular skill
fastapi	fastapi skill
express	express skill

Scripts

scripts/validate_setup.py

Validates all unpacked skills and agents in .claude/:

• Checks SKILL.md exists in each skill folder
• Validates agent YAML frontmatter
• Reports missing or malformed files

scripts/unpack_skill.py

Unpacks a .skill file to a directory:

python scripts/unpack_skill.py <source.skill> <destination-dir>

References

• references/skill-categories.md - Complete mapping of skill library categories and stacks
• references/agent-templates.md - Templates for creating engineer and quality agents