Code Explainer

When to Use

Use when asked to explain, walk through, or onboard onto a repo or codebase.

Shares the explainer skill's style rules — minimal text, diagrams over prose for structure/flow, concrete over abstract, no meta-phrasing. This file adds what's specific to explaining a whole repo.

Before writing anything: actually read it

Don't infer architecture from file/folder names alone. Open the entry point (main.go, cmd/), the README if one exists, and skim tests — tests often state intent more plainly than the implementation does. Guessing at structure from names and being wrong wastes more of his time than the extra minute of reading.

Procedure

1. Pass 1 — purpose. 1-2 plain sentences on what the repo is and who it's for. No code, no file names, no architecture yet.
2. Pass 2 — architecture. Entry point, layout, data flow, key components and how they connect. Draw this — a directory tree for layout, an arrow chain for data flow/request path (main → router → handler → store), not a paragraph narrating each hop.
3. Pass 3 — details with why. Non-obvious choices and what breaks if changed. For each key component, also flag: what could go wrong here (the failure modes this code has to guard against — a race, a nil, a partial write, a retry storm) and whether it could be simpler (over-engineered abstraction, unnecessary layer, a stdlib function reinventing itself) — call it out plainly if so, don't just describe the code as-is uncritically. Skip explaining known stdlib/language concepts and known patterns he already uses (see coding-style skill) — go straight to what's actually specific to this repo.
4. Close with how to run/verify (build/test/run commands actually found in the repo, not assumed) and where to look next if he wants to go deeper into one part.