菜单
Use when starting any conversation - establishes how arkhe skills bootstrap on Claude Code, Antigravity CLI (agy) / Gemini CLI, and Codex CLI, and maps Claude-only tools (AskUserQuestion, TaskCreate, EnterPlanMode, the Skill tool, the Agent tool) to their cross-platform equivalents
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/joaquimscosta/arkhe-claude-plugins --skill using-arkhe-skills命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Transform vague inputs into precision-optimized AI prompts for Claude, ChatGPT, Gemini, or other LLMs. Use when user mentions "optimize prompt", "improve prompt", "lyra", "prompt engineering", or needs help crafting effective AI prompts.
Validate plugin content quality before release. Run skill-validator across plugins, verify marketplace.json completeness, check plugin.json version consistency, and detect broken documentation cross-references. Use when preparing a release, before running release.sh, or when user mentions "release check", "pre-release", "plugin-release-checker", or "release validation".
Scan a project to detect configured quality and testing tools across JVM (Gradle/Maven), Android (AGP/Compose/KMP), Node.js/TypeScript, and Python ecosystems. Cross-reference against research-backed recommendations and assist with setup. Auto-detects project type(s) including monorepos with mixed ecosystems. Use when user asks to "audit tooling", "recommend tools", "quality stack", "what tools am I missing", "setup eslint", "setup detekt", "add coverage", "add ruff", "configure CI quality pipeline", "scan project tools", "tooling audit", "android tooling", "android quality", "compose testing", "kmp testing", or "screenshot testing".
基于 SOC 职业分类
正在显示 SKILL.md
| name | using-arkhe-skills |
| description | Use when starting any conversation - establishes how arkhe skills bootstrap on Claude Code, Antigravity CLI (agy) / Gemini CLI, and Codex CLI, and maps Claude-only tools (AskUserQuestion, TaskCreate, EnterPlanMode, the Skill tool, the Agent tool) to their cross-platform equivalents |
arkhe plugins ship the same SKILL.md files across CLI harnesses. This bootstrap establishes how to access skills and how to translate platform tool names into equivalent operations on Claude Code, Antigravity CLI (agy), Gemini CLI, and Codex CLI.
Skill tool. When invoked, the skill content is loaded and presented to follow directly.agy): Skill metadata (name + description) is auto-loaded into the system prompt at session start. When triggers fire, use view_file to load skills/<skill-name>/SKILL.md from the plugin directory, then follow its instructions.read_file to load skills/<skill-name>/SKILL.md.AGENTS.md at install time. Follow the skill body directly when a trigger phrase fires.arkhe SKILL.md files reference core agent primitives. On each platform, substitute the equivalent operation from this table.
| Primitive / Claude tool | Antigravity CLI (agy) equivalent | Gemini CLI (Legacy) equivalent | Codex CLI equivalent |
|---|---|---|---|
AskUserQuestion | ask_question (native interactive modal with chip UI options) | Ask a plain-text question with numbered choices and await reply. | Plain-text question with numbered choices. |
TaskCreate / TaskUpdate | manage_task or inline TODO list rendered on each turn | Inline TODO list rendered on each turn | Inline TODO list rendered on each turn |
EnterPlanMode / ExitPlanMode | Native --mode plan / accept-edits mode | Announce plan mode and gate writes on approval | Announce plan mode and gate writes on approval |
Skill tool (Skill) | view_file on skills/<name>/SKILL.md | read_file on skills/<name>/SKILL.md | Skill content inlined in AGENTS.md |
Agent tool (subagent_type) | invoke_subagent / define_subagent (native background/concurrent subagent execution) | Inline agent prompt into current session | Inline agent prompt into current session |
Commands ship from the same plugins/<plugin>/commands/<name>.md source on every platform, but the slash-command syntax differs:
| Platform | Form | Example |
|---|---|---|
| Claude Code | /<plugin>:<command> (plugin-namespaced) | /core:think |
Antigravity CLI (agy) / Gemini CLI | /<command> (or agy plugin commands) | /think |
| Codex CLI | Trigger phrase in chat (no slash) | "think through whether to use Postgres or Redis" |
When this skill or other arkhe docs refer to core:think, core:debug, etc., resolve them through this table — same command, platform-specific invocation.
Six arkhe commands dispatch subagents on Claude Code and Antigravity CLI (agy). On legacy Gemini and Codex CLI, they run inline with the agent's prompt collapsed into the command body.
The affected commands:
core:debug (with --deep)core:thinkcore:researchcore:double-check (with --deep)spring-boot:spring-reviewspring-boot:verify-upgradeIf CLAUDE.md, GEMINI.md, or AGENTS.md says "do not use TDD" and a skill says "always use TDD," follow the user's instructions. The user is in control.
plugins/<plugin>/skills/<skill>/SKILL.md.claude-plugin/marketplace.jsonplugins/<plugin>/plugin.json or .gemini-extensions/<plugin>/plugin.json (skills directory is a symlink to the canonical source).codex-marketplace/<plugin>/AGENTS.md (skill index synthesized at build time; full content via symlinked skills/)Skill bodies are not duplicated across platforms. One file, cross-platform support.