Codebase Onboarding

Skill — Codebase Onboarding

When this skill activates

When entering a new or unfamiliar repository for the first time, when asked to summarize or map a codebase, or when needing to build a mental model before making changes. Also activates when explicitly asked to generate an onboarding report or learning path.

Mandatory actions when this skill is active

Execute the following auto-summarization pipeline in order:

Step 1 — Detect Stack

Identify the technology stack by examining:

• Language(s): file extensions, shebang lines, language-specific config files
• Framework(s): package.json dependencies, requirements.txt, go.mod, Cargo.toml
• Build system: Makefile, webpack, vite, turbopack, gradle, cargo, mix
• Test framework: jest, pytest, vitest, go test, RSpec, xUnit
• Package manager: npm, yarn, pnpm, pip, poetry, cargo, go modules
• Infrastructure: Docker, Kubernetes, Terraform, CDK, serverless configs

Step 2 — Find Entry Points

Locate all entry points into the application:

• Main files (main.ts, index.ts, app.py, main.go, Program.cs)
• CLI entry points (bin/ directory, package.json bin field)
• Route definitions (router files, controller registrations)
• Event handlers (message consumers, cron jobs, webhook handlers)
• Worker/queue processors
• Migration entry points (seed files, migration runners)

Step 3 — Build Module Dependency Graph

Create a graph where:

• Nodes = modules, packages, or major source directories
• Edges = imports, function calls, event emissions between modules
• Identify circular dependencies (flag as architectural debt)
• Note external service dependencies (databases, APIs, queues)

Step 4 — Identify Hot Paths

Find the most critical files by measuring:

• Most imported: files that appear in the most import statements
• Most modified: files with the highest git commit frequency (last 90 days)
• Largest fan-out: files that import the most other modules
• Gateway files: files that bridge between major subsystems

Step 5 — Generate Learning Path

Create an ordered reading list that builds understanding progressively:

1. Configuration files (understand the shape of the project)
2. Entry points (understand how the system starts)
3. Core domain logic (understand what the system does)
4. Infrastructure/adapters (understand how it connects to the world)
5. Edge cases and error handling (understand failure modes)
6. Tests (understand expected behavior and invariants)

Step 6 — Output Report

Write ONBOARDING-REPORT.md to .planning/ containing:

• Stack summary (one-line per technology)
• Entry points list (file path + purpose)
• Dependency graph (ASCII or Mermaid format)
• Hot paths table (file, import count, commit frequency)
• Learning path (ordered reading list with file paths)
• Key architectural decisions or patterns observed
• Known debt or areas of concern

Self-check before task completion

Before marking a task done when this skill was active:

• Did I identify the stack correctly (language, framework, build, test)?
• Did I find ALL entry points (not just the obvious main file)?
• Did I build a dependency graph showing module relationships?
• Did I identify hot paths (most-imported and most-modified)?
• Did I create a usable learning path (ordered, progressive)?
• Did I output ONBOARDING-REPORT.md to .planning/?
• Is the report useful to someone with zero prior context?