name: project-detector description: > This skill should be used when the user asks to "detect project type", "analyze this project", "create project profile", "what tech stack is this", "scan project structure", "what database does this use", or when autonomous development starts in a new codebase. Detects package.json, pyproject.toml, go.mod, Cargo.toml to identify the technology stack. Also detects database providers (Supabase, Firebase, Prisma, etc.) and MCP availability.

Project Detection Skill

Analyze the current directory and create .claude/project-profile.yaml to enable autonomous development workflows.

Protocol

Step 1: Check Existing Profile

Check if .claude/project-profile.yaml exists:

• If exists and is less than 1 day old → read and summarize it
• Otherwise → create or update it

Step 2: Detect Package Manager and Language

Scan for project files:

ls package.json pyproject.toml go.mod Cargo.toml pom.xml 2>/dev/null

Identify the primary language based on which file exists.

Step 3: Detect Framework

Look for framework-specific config files:

Step 4: Extract Commands

Parse package.json scripts or detect standard commands:

• test - Testing command
• lint - Linting command
• typecheck - Type checking command
• build - Build command
• dev - Development server

Step 5: Detect Database Provider (NEW)

Check for database dependencies in package.json, pyproject.toml, or config files:

Also check for MCP availability:

• If Supabase detected → check if mcp__plugin_supabase_supabase__* tools are available
• MCP enables real-time schema validation

Step 6: Identify Directory Patterns

Scan directory structure to find:

• Component locations
• API route locations
• Test file locations
• Source code organization

Step 7: Write Profile

Write detected information to .claude/project-profile.yaml with structure:

project:
  name: ""
  type: ""  # frontend, backend, fullstack
stack:
  language: ""
  framework: ""
database:
  provider: ""  # supabase, firebase, prisma, drizzle, planetscale, neon, mongodb, postgres, mysql, sqlite, none
  orm: ""       # prisma, drizzle, typeorm, mongoose, kysely, none
  mcp_available: false  # true if MCP tools detected for this provider
  schema_source: ""     # mcp, prisma/schema.prisma, drizzle/schema.ts, manual
commands:
  test: ""
  lint: ""
  typecheck: ""
  build: ""
verification:
  required: []
  optional: []

Step 8: Summarize

Present detected configuration to the user for confirmation.

If database with MCP is detected, inform the user:

• Schema validation will use real-time MCP queries
• Type mismatches will be caught automatically
• RLS policies will be verified (Supabase)

Additional Resources

Reference Files

For detailed detection patterns:

• references/supported-frameworks.md - Complete list of supported frameworks with detection patterns
• references/database-providers.md - Database provider detection patterns and MCP integration

Example Files

Working examples in examples/:

• project-profile-nextjs.yaml - Example profile for a Next.js project

Script Reference

Run the detection script directly:

${CLAUDE_PLUGIN_ROOT}/scripts/detect-project.sh

When NOT to Use This Skill

Do NOT use this skill when:

1. Profile already exists and is recent - Skip if less than 1 day old
2. Empty directory - No project to detect
3. Non-project directories - Home folders, system directories, etc.
4. User explicitly provides tech stack - Trust user input over detection
5. Monorepo sub-packages - Detect from root, not from individual packages

Quality Standards

1. ALWAYS check for existing profile before regenerating
2. NEVER overwrite user-customized profile fields without confirmation
3. ALWAYS detect database provider if dependencies exist
4. ALWAYS check for MCP availability when database detected
5. PRIORITIZE parsing package.json/pyproject.toml over guessing