mcpField
D
documentation

documentation

github.com/epicenterhqDevelopment4.8k188 installsUpdated 3w ago
Repository

About documentation

A writing skill for documenting codebases, APIs, and comments. It emphasizes explaining design rationale and boundaries rather than restating implementation details, with guidance tailored to folder READMEs and code documentation.

Key features

  1. 01Explain design rationale and boundaries
  2. 02Structure folder README mental models
  3. 03Write JSDoc around when and why
  4. 04Use comments to clarify non-obvious reasoning

Use cases

  1. 01Introduce a folder's purpose and edit boundaries
  2. 02Document when to use a public API
  3. 03Explain a non-obvious code workaround

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/epicenterhq/epicenter --skill documentation

After installing, type/documentationin Claude Code to confirm it loaded.

Details

From the project README

Documentation

Follow writing-voice for tone. For architecture walkthroughs, folder mental models, or newcomer explanations, lead with the owner, boundary, or flow; ground the model in code and name the likely misconception when it helps.

Documentation explains why, not what. Users can read code to see what it does. They need you to explain the reasoning.

Folder READMEs

Primary job: explain why this folder exists and the mental model.

Open with the useful model: what this folder owns, why it exists, and what boundary future edits should preserve.

Can Include
  • ASCII art diagrams for complex relationships
  • Overview of key exports or entry points
  • Brief file descriptions IF they add context beyond the filename
  • Relationships to other folders
Avoid
  • Exhaustive file listings that just duplicate ls
  • Descriptions that repeat the filename ("auth.ts - authentication")
  • Implementation details better expressed in code
Good
# Converters

Transform field schemas into format-specific representations.

```
┌─────────────┐     ┌──────────────┐
│ Field Schema│────▶│  to-arktype  │────▶ Runtime validation
└─────────────┘     ├──────────────┤
                    │  to-drizzle  │────▶ SQLite columns
                    └──────────────┘
```

Field schemas are pure JSON Schema objects with `x-component` hints. Each converter takes the same input and produces output for a specific consumer.
Bad
# Converters

- `to-arktype.ts` - Converts to ArkType
- `to-drizzle.ts` - Converts to Drizzle
- `index.ts` - Exports

The bad example just lists files without explaining the pattern or when to add new converters.

JSDoc Comments

JSDoc explains when and why to use something, not just what it does.

Good
/**
 * Get all table helpers as an array.
 *
 * Useful for providers and indexes that need to iterate over all tables.
 * Returns only the table helpers, excluding utility methods like `clearAll`.
 *
 * @example
 * ```typescript
 * for (const table of tables.defined()) {
 *   console.log(table.name, table.count());
 * }
 * ```
 */
defined() { ... }
Bad
/** Returns all table helpers as an array. */
defined() { ... }
Rules
  • Include @example blocks with realistic usage
  • Explain WHEN to use it, not just WHAT it does
  • Document non-obvious behavior or edge cases
  • Public APIs get detailed docs; internal helpers can be minimal

Code Comments

Comments explain why, not what.

Good
// Y.Doc clientIDs are random 32-bit integers, so we can't rely on ordering.
// Use timestamps from the entries themselves for deterministic sorting.
const sorted = entries.sort((a, b) => a.timestamp - b.timestamp);
Bad
// Sort the entries
const sorted = entries.sort((a, b) => a.timestamp - b.timestamp);
Rules
  • If the code is clear, don't comment it
  • Comment the "why" when it's not obvious
  • Comment workarounds with links to issues/docs
  • Delete commented-out code; that's what git is for

Comments

Markdown supported