CLAUDE.md

1# CLAUDE.md
2 
3This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
4 
5## Repository Overview
6 
7This is a best practices repository for Claude Code configuration, demonstrating patterns for skills, subagents, hooks, and commands. It serves as a reference implementation rather than an application codebase.
8 
9## Key Components
10 
11### Weather System (Example Workflow)
12A demonstration of two distinct skill patterns via the **Command → Agent → Skill** architecture:
13- `/weather-orchestrator` command (`.claude/commands/weather-orchestrator.md`): Entry point — asks user for C/F, invokes agent, then invokes SVG skill
14- `weather-agent` agent (`.claude/agents/weather-agent.md`): Fetches temperature using its preloaded `weather-fetcher` skill (agent skill pattern)
15- `weather-fetcher` skill (`.claude/skills/weather-fetcher/SKILL.md`): Preloaded into agent — instructions for fetching temperature from Open-Meteo
16- `weather-svg-creator` skill (`.claude/skills/weather-svg-creator/SKILL.md`): Skill — creates SVG weather card, writes `orchestration-workflow/weather.svg` and `orchestration-workflow/output.md`
17 
18Two skill patterns: agent skills (preloaded via `skills:` field) vs skills (invoked via `Skill` tool). See `orchestration-workflow/orchestration-workflow.md` for the complete flow diagram.
19 
20### Skill Definition Structure
21Skills in `.claude/skills/<name>/SKILL.md` use YAML frontmatter:
22- `name`: Display name and `/slash-command` (defaults to directory name)
23- `description`: When to invoke (recommended for auto-discovery)
24- `argument-hint`: Autocomplete hint (e.g., `[issue-number]`)
25- `disable-model-invocation`: Set `true` to prevent automatic invocation
26- `user-invocable`: Set `false` to hide from `/` menu (background knowledge only)
27- `allowed-tools`: Tools allowed without permission prompts when skill is active
28- `model`: Model to use when skill is active
29- `context`: Set to `fork` to run in isolated subagent context
30- `agent`: Subagent type for `context: fork` (default: `general-purpose`)
31- `hooks`: Lifecycle hooks scoped to this skill
32 
33### Presentation System
34See `.claude/rules/presentation.md` — presentation work is delegated per-presentation to `presentation-vibe-coding` (for `presentation/vibe-coding-to-agentic-engineering/`) or `presentation-claude-gemini` (for `presentation/2026-04-25-gdg-kolachi-cli-claude-code-gemini/`).
35 
36### Hooks System
37Cross-platform sound notification system in `.claude/hooks/`:
38- `scripts/hooks.py`: Main handler for Claude Code hook events
39- `config/hooks-config.json`: Shared team configuration
40- `config/hooks-config.local.json`: Personal overrides (git-ignored)
41- `sounds/`: Audio files organized by hook event (generated via ElevenLabs TTS)
42 
43Hook events configured in `.claude/settings.json`: PreToolUse, PostToolUse, UserPromptSubmit, Notification, Stop, SubagentStart, SubagentStop, PreCompact, SessionStart, SessionEnd, Setup, PermissionRequest, TeammateIdle, TaskCompleted, ConfigChange.
44 
45Special handling: git commits trigger `pretooluse-git-committing` sound.
46 
47## Critical Patterns
48 
49### Subagent Orchestration
50Subagents **cannot** invoke other subagents via bash commands. Use the Agent tool (renamed from Task in v2.1.63; `Task(...)` still works as an alias):
51```
52Agent(subagent_type="agent-name", description="...", prompt="...", model="haiku")
53```
54 
55Be explicit about tool usage in subagent definitions. Avoid vague terms like "launch" that could be misinterpreted as bash commands.
56 
57### Subagent Definition Structure
58Subagents in `.claude/agents/*.md` use YAML frontmatter:
59- `name`: Subagent identifier
60- `description`: When to invoke (use "PROACTIVELY" for auto-invocation)
61- `tools`: Comma-separated allowlist of tools (inherits all if omitted). Supports `Agent(agent_type)` syntax
62- `disallowedTools`: Tools to deny, removed from inherited or specified list
63- `model`: Model alias: `haiku`, `sonnet`, `opus`, or `inherit` (default: `inherit`)
64- `permissionMode`: Permission mode (e.g., `"acceptEdits"`, `"plan"`, `"bypassPermissions"`)
65- `maxTurns`: Maximum agentic turns before the subagent stops
66- `skills`: List of skill names to preload into agent context
67- `mcpServers`: MCP servers for this subagent (server names or inline configs)
68- `hooks`: Lifecycle hooks scoped to this subagent (all hook events are supported; `PreToolUse`, `PostToolUse`, and `Stop` are the most common)
69- `memory`: Persistent memory scope — `user`, `project`, or `local` (see `reports/claude-agent-memory.md`)
70- `background`: Set to `true` to always run as a background task
71- `effort`: Effort level override: `low`, `medium`, `high`, `max` (default: inherits from session)
72- `isolation`: Set to `"worktree"` to run in a temporary git worktree
73- `color`: CLI output color for visual distinction
74 
75### Configuration Hierarchy
761. **Managed** (`managed-settings.json` / MDM plist / Registry): Organization-enforced, cannot be overridden
772. Command line arguments: Single-session overrides
783. `.claude/settings.local.json`: Personal project settings (git-ignored)
794. `.claude/settings.json`: Team-shared settings
805. `~/.claude/settings.json`: Global personal defaults
816. `hooks-config.local.json` overrides `hooks-config.json`
82 
83### Disable Hooks
84Set `"disableAllHooks": true` in `.claude/settings.local.json`, or disable individual hooks in `hooks-config.json`.
85 
86## Answering Best Practice Questions
87 
88When the user asks a Claude Code best practice question, **always search this repo first** (`best-practice/`, `reports/`, `tips/`, `implementation/`, and `README.md`) before relying on training knowledge or external sources. This repo is the authoritative source — only fall back to external docs or web search if the answer is not found here.
89 
90## Workflow Best Practices
91 
92From experience with this repository:
93 
94- Keep CLAUDE.md under 200 lines per file for reliable adherence
95- `.claude/rules/*.md` with `paths:` YAML frontmatter are lazy-loaded only when Claude touches matching files; without frontmatter they load into every session like CLAUDE.md
96- Use commands for workflows instead of standalone agents
97- Create feature-specific subagents with skills (progressive disclosure) rather than general-purpose agents
98- Perform manual `/compact` at ~50% context usage
99- Start with plan mode for complex tasks
100- Use human-gated task list workflow for multi-step tasks
101- Break subtasks small enough to complete in under 50% context
102 
103### Debugging Tips
104 
105- Use `/doctor` for diagnostics
106- Run long-running terminal commands as background tasks for better log visibility
107- Use browser automation MCPs (Claude in Chrome, Playwright, Chrome DevTools) for Claude to inspect console logs
108- Provide screenshots when reporting visual issues
109 
110## Git Commit Rules
111 
112When committing changes, **create separate commits per file**. Do NOT bundle multiple file changes into a single commit. Each file gets its own commit with a descriptive message specific to that file's changes.
113 
114For example, if `README.md`, `best-practice/claude-subagents.md`, and a skill file all changed:
115- Commit 1: `git add README.md` → commit with README-specific message
116- Commit 2: `git add best-practice/claude-subagents.md` → commit with subagents-doc-specific message
117- Commit 3: `git add .claude/skills/weather-fetcher/SKILL.md` → commit with skill-specific message
118 
119This makes the git history cleaner and easier to review, revert, or cherry-pick individual changes.
120 
121## Documentation
122 
123See `.claude/rules/markdown-docs.md` for documentation standards. Key docs:
124- `best-practice/claude-subagents.md`: Subagent frontmatter, hooks, and repository agents
125- `best-practice/claude-commands.md`: Slash command patterns and built-in command reference
126- `orchestration-workflow/orchestration-workflow.md`: Weather system flow diagram
127