# How

> Walk through how a subsystem works — a fast explorer maps the territory, an explainer writes the layered walkthrough. Use for "how does X work", onboarding on unfamiliar code, or before changing a subsystem you cannot yet explain.

- Skill: `tuyv/how` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx get-skill add tuyv/how`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tuyv/how/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: tuyv (https://skillmd.com/u/tuyv)
- Updated: 2026-11-14
- Page: https://skillmd.com/skills/tuyv/how

---


# How

Explain how a subsystem works, layered so each layer stands alone. Two roles,
two passes:

1. **Explore** (read-only, fast): dispatch `oh-my-claudecode:explore` to map
   the subsystem — entry points, main flows, data shapes, state, external
   calls, and the files that own each. Explorer returns a map with file:line
   citations, not prose.
2. **Explain** (lead): write the walkthrough from the map, reading the cited
   code yourself where the claim is load-bearing.

## Output shape

- **One paragraph**: what the subsystem does and for whom, no jargon.
- **Walkthrough**: the main path end to end, step by step, each step citing
  `file:line`. Diverge into how errors and retries are handled at the point
  they occur.
- **Edge cases and invariants**: what must stay true for the walkthrough to
  hold, and what happens when it does not.
- **Boundaries**: what this subsystem deliberately does not do.

## Rules

- Every behavioral claim cites code you or the explorer read this session;
  an uncited claim is labeled a guess.
- The walkthrough explains the code as it is, not as it should be; design
  critique belongs to `why` and `review`.
- Size the depth to the ask: a routing question gets one paragraph plus the
  relevant slice, not the full walkthrough.

