multiAI Summary Pending

llm-doc-writer

Write token-efficient documentation for LLM context. Use when creating CLAUDE.md, README, technical docs, agent instructions, or any documentation consumed by AI assistants.

231 stars

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/llm-doc-writer/SKILL.md --create-dirs "https://raw.githubusercontent.com/aiskillstore/marketplace/main/skills/aceek/llm-doc-writer/SKILL.md"

Manual Installation

  1. Download SKILL.md from GitHub
  2. Place it in .claude/skills/llm-doc-writer/SKILL.md inside your project
  3. Restart your AI agent — it will auto-discover the skill

How llm-doc-writer Compares

Feature / Agentllm-doc-writerStandard Approach
Platform SupportmultiLimited / Varies
Context Awareness High Baseline
Installation ComplexityUnknownN/A

Frequently Asked Questions

What does this skill do?

Write token-efficient documentation for LLM context. Use when creating CLAUDE.md, README, technical docs, agent instructions, or any documentation consumed by AI assistants.

Which AI agents support this skill?

This skill is compatible with multi.

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

SKILL.md Source

# LLM-Optimized Documentation

## Core Principles

| Principle | Rule |
|-----------|------|
| **Density** | Max info, min tokens |
| **Format** | Tables > prose, bullets > paragraphs |
| **No fluff** | Ban filler words (see list below) |
| **Show don't tell** | Code examples > explanations |
| **Progressive disclosure** | TOC + separate files for details |

## Banned Patterns

```
# Filler words - NEVER use
simplement, il suffit de, en fait, basically, just, simply,
it's important to note, as mentioned, obviously, of course,
please note that, keep in mind, remember that

# Redundant structures - NEVER use
"This file contains..." (obvious from filename)
"In this section we will..." (just do it)
"The following example shows..." (just show it)
```

## Format Rules

### Tables over Prose

```markdown
# BAD - 45 tokens
The system supports three modes: development mode which
enables hot reload, production mode which optimizes for
performance, and test mode which mocks external services.

# GOOD - 20 tokens
| Mode | Behavior |
|------|----------|
| dev | Hot reload |
| prod | Optimized |
| test | Mocked services |
```

### Bullets over Paragraphs

```markdown
# BAD - Narrative
To run the project, first ensure Node.js is installed,
then install dependencies with npm install, and finally
start the dev server with npm run dev.

# GOOD - Scannable
## Run
1. Requires: Node.js 18+
2. `npm install`
3. `npm run dev`
```

### Code over Explanation

```markdown
# BAD - Explaining
To create a new user, call the createUser function
with an object containing name and email properties.

# GOOD - Showing
```ts
createUser({ name: "Jo", email: "jo@x.com" })
```
```

## Structure Template

For CLAUDE.md / agent docs:

```markdown
# Project Name

## Stack
- Frontend: React/Vite
- Backend: FastAPI
- DB: PostgreSQL

## Commands
| Action | Command |
|--------|---------|
| Dev | `npm run dev` |
| Test | `npm test` |
| Build | `npm run build` |

## Architecture
[Brief description, link to detailed docs if needed]

## Conventions
- [Rule 1]
- [Rule 2]
```

## Progressive Disclosure

For long docs (>200 lines):

```markdown
# Main Doc

## Quick Reference
[Essential info here]

## Details
See [ARCHITECTURE.md](ARCHITECTURE.md)
See [API.md](API.md)
```

## Self-Check

Before finalizing, verify:

- [ ] No filler words?
- [ ] Tables used where possible?
- [ ] Bullet points, not paragraphs?
- [ ] Examples over explanations?
- [ ] < 500 lines (or split)?
- [ ] No redundant info?

## Examples

See [patterns.md](patterns.md) for before/after examples.