docs(CLAUDE): condense guidance and add hard rules + token efficiency

This commit is contained in:
Luong NGUYEN
2026-05-02 00:41:12 +02:00
parent d2ead3ba16
commit f925478d78
+41 -136
View File
@@ -1,157 +1,62 @@
# CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Tutorial repo. Output is markdown in numbered modules `01-` through `10-`, not an app. Scripts in `scripts/` exist only to validate docs and build the EPUB.
## Project Overview
See also `.claude/CLAUDE.md` for stack/commands and `STYLE_GUIDE.md` for lesson structure.
Claude How To is a tutorial repository for Claude Code features. This is **documentation-as-code** — the primary output is markdown files organized into numbered learning modules, not an executable application.
**Architecture**: Each module (01-10) covers a specific Claude Code feature with copy-paste templates, Mermaid diagrams, and examples. The build system validates documentation quality and generates an EPUB ebook.
## Common Commands
### Pre-commit Quality Checks
All documentation must pass four quality checks before commits (these run automatically via pre-commit hooks):
## Critical commands
```bash
# Install pre-commit hooks (runs on every commit)
pre-commit install
# Run all checks manually
# Quality gate (also runs on commit via pre-commit hooks)
pre-commit run --all-files
```
The four checks are:
1. **markdown-lint** — Markdown structure and formatting via `markdownlint`
2. **cross-references** — Internal links, anchors, code fence syntax (Python script)
3. **mermaid-syntax** — Validates all Mermaid diagrams parse correctly (Python script)
4. **link-check** — External URLs are reachable (Python script)
5. **build-epub** — EPUB generates without errors (on `.md` changes)
### Development Environment Setup
```bash
# Install uv (Python package manager)
pip install uv
# Create virtual environment and install Python dependencies
uv venv
source .venv/bin/activate
uv pip install -r scripts/requirements-dev.txt
# Install Node.js tools (markdown linter and Mermaid validator)
npm install -g markdownlint-cli
npm install -g @mermaid-js/mermaid-cli
# Install pre-commit hooks
uv pip install pre-commit
pre-commit install
```
### Testing
Python scripts in `scripts/` have unit tests:
```bash
# Run all tests
# Tests
pytest scripts/tests/ -v
# Run with coverage
pytest scripts/tests/ -v --cov=scripts --cov-report=html
# Run specific test
pytest scripts/tests/test_build_epub.py -v
```
### Code Quality
```bash
# Lint and format Python code
ruff check scripts/
ruff format scripts/
# Security scan
bandit -c scripts/pyproject.toml -r scripts/ --exclude scripts/tests/
# Type checking
mypy scripts/ --ignore-missing-imports
```
### EPUB Build
```bash
# Generate ebook (renders Mermaid diagrams via Kroki.io API)
# EPUB build (calls Kroki.io API to render Mermaid — needs network)
uv run scripts/build_epub.py
# With options
uv run scripts/build_epub.py --verbose --output custom-name.epub --max-concurrent 5
# Python tooling
ruff check scripts/ && ruff format scripts/
mypy scripts/ --ignore-missing-imports
bandit -c scripts/pyproject.toml -r scripts/ --exclude scripts/tests/
```
## Directory Structure
Pre-commit runs 5 checks: markdown-lint, cross-references, mermaid-syntax, link-check, build-epub (on `.md` changes). All must pass.
```
├── 01-slash-commands/ # User-invoked shortcuts
├── 02-memory/ # Persistent context examples
├── 03-skills/ # Reusable capabilities
├── 04-subagents/ # Specialized AI assistants
├── 05-mcp/ # Model Context Protocol examples
├── 06-hooks/ # Event-driven automation
├── 07-plugins/ # Bundled features
├── 08-checkpoints/ # Session snapshots
├── 09-advanced-features/ # Planning, thinking, backgrounds
├── 10-cli/ # CLI reference
├── scripts/
│ ├── build_epub.py # EPUB generator (renders Mermaid via Kroki API)
│ ├── check_cross_references.py # Validates internal links
│ ├── check_links.py # Checks external URLs
│ ├── check_mermaid.py # Validates Mermaid syntax
│ └── tests/ # Unit tests for scripts
├── .pre-commit-config.yaml # Quality check definitions
└── README.md # Main guide (also module index)
```
## Architecture map
## Content Guidelines
- `01-``10-` — tutorial modules. **Numbered prefix = learning order**, not alphabetical. Do not reorganize.
- Each module: `README.md` + copy-paste templates (`.md`, `.json`, `.sh`).
- `scripts/` — utilities (EPUB builder, link/mermaid/cross-ref validators). Not the product.
- `02-memory/*.md` — CLAUDE.md templates users copy into their own projects. Don't confuse with this file.
- `openspec/` — spec-driven change proposals.
### Module Structure
Each numbered folder follows the pattern:
- **README.md** — Overview of the feature with examples
- **Example files** — Copy-paste templates (`.md` for commands, `.json` for configs, `.sh` for hooks)
- Files are organized by feature complexity and dependencies
## Hard rules
### Mermaid Diagrams
- All diagrams must parse successfully (checked by pre-commit hook)
- EPUB build renders diagrams via Kroki.io API (requires internet)
- Use Mermaid for flowcharts, sequence diagrams, and architecture visuals
- **YOU MUST NOT commit or push without explicit user request.**
- **YOU MUST NOT add `Co-Authored-By: Claude`** to any commit message.
- Always activate `.venv` before running Python scripts (check `venv/`, `.venv/`, `env/`).
- Internal links use **relative paths** (e.g. `01-slash-commands/README.md`); anchors use `#heading-name`.
- Code fences **must** declare a language (`bash`, `python`, `json`, …) — the cross-reference check fails otherwise.
- External URLs must be reachable and stable. No ephemeral links.
- Mermaid diagrams must parse (validated pre-commit). Broken EPUB build is usually invalid Mermaid or no network to Kroki.
- Commit format: `type(scope): subject` where `scope` matches the module folder (e.g. `feat(slash-commands):`, `docs(memory):`, `fix(README):`).
- Do not reorganize the `01-``10-` numbering. The order is the curriculum.
### Cross-References
- Use relative paths for internal links (e.g., `(01-slash-commands/README.md)`)
- Code fences must specify language (e.g., ` ```bash `, ` ```python `)
- Anchor links use `#heading-name` format
## Workflow preferences
### Link Validation
- External URLs must be reachable (checked by pre-commit hook)
- Avoid linking to ephemeral content
- Use permalinks where possible
- For lesson edits, follow `STYLE_GUIDE.md` for structure/naming/diagrams.
- Small fixes → minimal diff. Don't rewrite a section to fix a typo.
- When adding a module page: README + templates first, then update root `README.md` index and `LEARNING-ROADMAP.md` if order/timing changes.
- Tutorial > library: prioritize clear explanations and copy-paste examples over reusable abstractions.
- If a quality check fails, fix the underlying issue. Don't bypass with `--no-verify`.
## Key Architecture Points
1. **Numbered folders indicate learning order** — The 01-10 prefix represents the recommended sequence for learning Claude Code features. This numbering is intentional; do not reorganize alphabetically.
2. **Scripts are utilities, not the product** — The Python scripts in `scripts/` support documentation quality and EPUB generation. The actual content is in the numbered module folders.
3. **Pre-commit is the gatekeeper** — All four quality checks must pass before a PR is accepted. The CI pipeline runs these same checks as a second pass.
4. **Mermaid rendering requires network** — The EPUB build calls Kroki.io API to render diagrams. Build failures here are typically network issues or invalid Mermaid syntax.
5. **This is a tutorial, not a library** — When adding content, focus on clear explanations, copy-paste examples, and visual diagrams. The value is in teaching concepts, not providing reusable code.
## Commit Conventions
Follow conventional commit format:
- `feat(slash-commands): Add API documentation generator`
- `docs(memory): Improve personal preferences example`
- `fix(README): Correct table of contents link`
- `refactor(hooks): Simplify hook configuration examples`
Scope should match the folder name when applicable.
## Token Efficiency
- Never re-read files you just wrote or edited. You know the contents.
- Never re-run commands to "verify" unless the outcome was uncertain.
- Don't echo back large blocks of code or file contents unless asked.
- Batch related edits into single operations. Don't make 5 edits when 1 handles it.
- Skip confirmations like "I'll continue..." Just do it.
- If a task needs 1 tool call, don't use 3. Plan before acting.
- Do not summarize what you just did unless the result is ambiguous or you need additional input.