Bootstrap Skill

Goal

Transform any project — new or existing — into a fully configured Ground Zero workspace by scanning, analyzing, and generating project-specific configuration.

Instructions

Determine Project Type

First, check if the project has existing code:

• If the workspace contains only .agent/, GEMINI.md, docs/, and README.md → this is a NEW project
• If the workspace contains other source files, config files, or directories → this is an EXISTING project

For NEW Projects

1. Ask the orchestrator what they want to build. Ask about:

   • What is the project? (one sentence description)
   • What platform? (web app, mobile app, API, CLI tool, etc.)
   • Any technology preferences? (or should you recommend?)
   • Any constraints? (budget, timeline, must-use services)

2. Recommend a tech stack based on the answers. Explain each choice in plain English. Create an ADR for the tech stack decision.

3. Generate project-specific files:

   • A rule file in .agent/rules/ for the chosen code style
   • A rule file for the chosen architecture conventions
   • Skills for: build, test, run/serve, lint (based on the tech stack)
   • Update .agent/knowledge/tech-stack.md
   • Update .agent/knowledge/project-map.md with the planned structure
   • Add framework-specific terms to .agent/knowledge/glossary.md
   • Log all generated items to .agent/knowledge/evolution-log.md

4. Report to orchestrator: Plain-English summary of everything generated and why.

For EXISTING Projects

1. Scan the directory tree — list all directories and file types present

2. Detect the tech stack — read the checklist in resources/scan-checklist.md and the signatures in resources/tech-stack-signatures.md. Look for:

   • Package manager files (package.json, requirements.txt, Cargo.toml, go.mod, etc.)
   • Framework config files (next.config.js, vite.config.ts, django settings, etc.)
   • CI/CD files (.github/workflows/, Dockerfile, etc.)
   • Test configuration (jest.config, pytest.ini, etc.)

3. Sample code files — read 10-15 representative source files to detect:

   • Coding style (tabs/spaces, naming conventions, semicolons, quotes)
   • Architecture patterns (MVC, component-based, microservices, etc.)
   • Common patterns (error handling style, logging, validation approach)
   • Test patterns (test file naming, assertion style)

4. Read existing documentation — README, CONTRIBUTING, any docs/ folder

5. Generate project-specific files:

   • Rules matching the detected code style
   • Rules matching the detected architecture conventions
   • Skills for detected build/test/lint/deploy commands
   • Update .agent/knowledge/tech-stack.md with detected technologies
   • Update .agent/knowledge/project-map.md with directory descriptions
   • Add project-specific terms to .agent/knowledge/glossary.md
   • Log all generated items to .agent/knowledge/evolution-log.md

6. Report to orchestrator: Plain-English summary of what was found and what was generated.

Constraints

• Never modify existing source code during bootstrap — only create .agent/ configuration files
• Always ask the orchestrator for approval before generating rules or skills
• If the codebase is very large (>500 files), do a representative sample rather than reading everything
• State your confidence level in detected patterns ("I am 90% sure this uses React because...")