| name | codebase-onboarding |
| description | Systematically analyze unfamiliar codebases through reconnaissance, architecture mapping, convention detection, and artifact generation (onboarding guide + CLAUDE.md). |
| origin | ECC |
Codebase Onboarding
Systematically analyze unfamiliar codebases to produce actionable onboarding artifacts.
When to Use
- Joining a new project or team
- Onboarding to an unfamiliar codebase
- Generating project documentation for new contributors
- Creating a starter CLAUDE.md for AI-assisted development
Phases
Phase 1: Reconnaissance
Gather initial signals without reading every file:
- Detect package manifests and frameworks
- Identify entry points and directory structure
- Find config files and test setup
Phase 2: Architecture Mapping
- Identify tech stack and architecture pattern (monolith vs. microservices)
- Map key directories and their purposes
- Trace data flow from request entry to database
Phase 3: Convention Detection
Discover project conventions from actual code patterns:
- Naming patterns and code style
- Error handling approaches
- Git workflow and commit conventions
Phase 4: Generate Artifacts
Produce two outputs:
Onboarding Guide (scannable in ~2 minutes):
- Tech stack table
- Architecture overview
- Key entry points
- Directory mapping
- Request lifecycle
- Conventions summary
- Common task commands
- "Where to look" reference table
CLAUDE.md (under 100 lines):
- Project-specific instructions for code style, testing, build commands, and structure
Key Principles
- Use targeted searches (Glob, Grep) rather than reading all files
- Trust actual code patterns over config declarations
- Enhance existing CLAUDE.md files rather than replacing them
- Flag unknowns explicitly -- avoid confident guesses
- Keep outputs concise and scannable, not exhaustive