Onboard

/onboard — map the codebase

Goal: understand a codebase enough to work in it, cheaply — without reading everything.

Do this

1. Structure first — read the manifests (package.json / pyproject / README) and the top-level tree. Identify the stack, scripts, and entry points.
2. Use the graph / mgrep (or Grep/Glob if those are unavailable), not brute-force file reads: find the main modules, the request/data flow, and the seams (where layers meet).
3. Conventions — note the patterns the repo already uses (naming, error handling, state management, test layout). Future work must match the grain.
4. Hot spots — where is the core logic, where are the tests, where does config live, what's the build/run/test command.

Output — a lean codemap

## Stack & run/build/test commands
## Entry points & main flow
## Module map (dir → responsibility)
## Conventions to follow
## Where to add things (tests, config, features)

Save it to the session file so later phases don't re-explore. Keep it to one screen.