菜单
Guide for creating a new Claude Code skill in this project. Use when asked to convert knowledge into a skill, create a slash command, or add a new skill file.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
How to use @owlmeans/native-client — the React Native app bootstrap package. Provides makeContext() to wire context/auth/permissions/router/db, AppConfig for multi-environment service routing, and useContext() hook. Auto-invoked when importing from native-client or setting up a React Native OwlMeans app.
How to use @owlmeans/native-db — React Native AsyncStorage-backed ClientDb adapter. Use makeNativeDbService() to create a multi-store persistent key/value service, or appendNativeDbService() to register it on a context. Auto-invoked when importing from native-db or wiring local storage in a React Native app.
How to use @owlmeans/native-panel — React Native Material Design 3 panel/UI component library built on react-native-paper. Use PanelApp as the root wrapper; re-exports makeContext, common panel helpers, and form/modal/list components. Auto-invoked when importing from native-panel or building React Native admin/dashboard UIs.
How to use @owlmeans/native-router — React Router Native adapter implementing the @owlmeans/router contract. Use appendNativeRouter() to register a memory-router-based router service on a context. Auto-invoked when importing from native-router or wiring navigation in a React Native OwlMeans app.
How to use @owlmeans/native-client — the React Native app bootstrap package. Provides makeContext() to wire context/auth/permissions/router/db, AppConfig for multi-environment service routing, and useContext() hook. Auto-invoked when importing from native-client or setting up a React Native OwlMeans app.
How to use @owlmeans/native-db — React Native AsyncStorage-backed ClientDb adapter. Use makeNativeDbService() to create a multi-store persistent key/value service, or appendNativeDbService() to register it on a context. Auto-invoked when importing from native-db or wiring local storage in a React Native app.
| name | create-skill |
| description | Guide for creating a new Claude Code skill in this project. Use when asked to convert knowledge into a skill, create a slash command, or add a new skill file. |
Skills live in .claude/skills/<skill-name>/SKILL.md. Each skill is a directory.
.claude/skills/<skill-name>/
├── SKILL.md # Required — entrypoint with frontmatter + instructions
├── reference.md # Optional — detailed reference docs
├── examples.md # Optional — usage examples
└── scripts/ # Optional — bundled shell scripts the skill runs
└── <name>.sh
Some skills bundle shell scripts that agents run as part of their procedure. This requires two locations for dual-tool support:
| Location | Purpose |
|---|---|
.claude/skills/<name>/scripts/<script>.sh | Claude Code — loaded relative to the skill |
.github/agent-scripts/<script>.sh | GitHub Copilot — accessible from repo root |
Both copies must remain byte-identical. When updating a script, update both locations simultaneously. Document this sync requirement in the SKILL.md.
When creating a skill with bundled scripts:
.claude/skills/<name>/scripts/<script>.sh.github/agent-scripts/<script>.shallowed-tools: Bash(sh *) in frontmatter so Claude can run the script without prompting---
name: skill-name # becomes the /slash-command (lowercase, hyphens)
description: What it does and when to use it # guides auto-invocation by Claude
allowed-tools: Bash(bun *) Read # tools usable without per-call approval
disable-model-invocation: true # set true for side-effect tasks (deploy, commit)
user-invocable: false # set false for background knowledge only
---
| Field | Use when |
|---|---|
description | Always — primary signal for when Claude auto-invokes |
allowed-tools | Skill needs to run specific commands without prompting |
disable-model-invocation: true | Skill has side effects (git push, deploy, etc.) — user must invoke manually |
user-invocable: false | Skill is background knowledge, not an action — hide from / menu |
argument-hint | Skill takes arguments — show hint in autocomplete |
context: fork | Skill should run in isolated subagent (long research, exploration) |
$ARGUMENTS — everything typed after /skill-name$0, $1, $2 — individual arguments by index!`command` — shell command, executes before Claude sees the skill (output replaces it)Example with dynamic context:
Current branch: !`git rev-parse --abbrev-ref HEAD`
| Scenario | Approach |
|---|---|
| Project-specific CLI/build knowledge | Skill, no disable-model-invocation (let Claude auto-use) |
| Side-effect workflow (commit, deploy) | disable-model-invocation: true |
| Background reference only | user-invocable: false |
| Runs long searches or exploration | context: fork, agent: Explore |
.claude/<topic>.md file the skill replaces.claude/memory/MEMORY.md to mention the skill existsCLAUDE.md Additional Context section if the old file was referenced there/skill-name in Claude Code.claude/memory/*.md): facts about decisions, history, or one-off project state that doesn't fit a repeatable pattern