菜单
Run Gemini CLI for AI queries. Use when user asks to "run/ask/use gemini", compare Claude vs Gemini, or delegate tasks to Gemini.
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/georgekhananaev/claude-skills-vault --skill gemini-cli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。
正在显示 SKILL.md
| name | gemini-cli |
| description | Run Gemini CLI for AI queries. Use when user asks to "run/ask/use gemini", compare Claude vs Gemini, or delegate tasks to Gemini. |
Interact w/ Google's Gemini CLI locally. Run queries, get responses, compare outputs.
Gemini CLI must be installed & configured:
gemini & sign in w/ Google accountgemini --versionIMPORTANT: Use -p flag for non-interactive (headless) one-shot queries. Without -p, gemini opens interactive mode which hangs in automation.
# One-shot query (MUST use -p for non-interactive)
gemini -p "Your prompt"
# Specific model
gemini -p "prompt" -m gemini-3-pro
# JSON output
gemini -p "prompt" -o json
# Auto-approve tool use (-y is deprecated)
gemini --approval-mode=yolo -p "prompt"
# File analysis
cat file.txt | gemini -p "Analyze this"
| Alias | Use case |
|---|---|
auto | DEFAULT — routes per prompt (simple → Flash, complex → 3 Pro) |
pro | Force Gemini 3 Pro (best all-round) |
flash / flash-lite | Fast, lightweight tasks (3.5 Flash GA rolling out as auto default) |
Omit -m to use auto routing. Aliases beat pinned model ids — they track
upstream model swaps. Pin a full id (e.g. gemini-3-pro) only when the user
names one.
| Flag | Desc |
|---|---|
-p | Required for headless. Non-interactive prompt |
-m | Model selection |
-o | Output: text/json/stream-json |
--approval-mode=yolo | Auto-approve all actions (-y is deprecated) |
-d | Debug mode |
-s | Sandbox mode |
-r | Resume session |
-i | Execute prompt then continue interactive |
gemini -p "prompt"-p for automation/one-shot queries-o json for parsingShapes how every reply to the user gets written - plain everyday words instead of corporate or AI-sounding phrasing, real sentences instead of bullet-heavy formatting, no forced dash punctuation, and at most a rare bit of dry humor when it genuinely fits. Use this for every conversational response written to the user in this project, not only on special request - check it before replying, the same way a person reads over their own message before hitting send.
Safety-first Firebase CLI (firebase-tools v15) skill for full project control — deploy, Hosting (sites/channels/rollback), Cloud Functions (+secrets), Firestore (databases/indexes/backups/delete), Realtime Database, Auth import/export, Remote Config, App Distribution, App Hosting, Extensions, Data Connect, Emulator Suite & MCP server. Classifies every command by risk tier via a deterministic classifier script and gates destructive/irreversible/cost-incurring ops behind AskUserQuestion confirmation; enforces the --non-interactive/--force contract so nothing hangs and nothing is auto-confirmed. Wrong-project preflight prevents deploying to prod by accident. Ships a 3-level self-test (static classifier battery, live read-only, guarded live-write w/ cleanup). Use when running, planning, or debugging any `firebase` command.
Run OpenAI Codex CLI for coding tasks, implementation, reviews, and second-opinion audits with mandatory task-based routing across GPT-5.6-or-newer models. Use when a user asks to run, ask, or use Codex; says "codex prompt"; wants a Codex code/logic/plan audit; or wants Claude to delegate work to OpenAI models. Inspect the live Codex model catalog, explicitly pin an eligible model and reasoning effort on every invocation, route clear high-volume work to Luna, everyday work to Terra, and difficult or high-value work to Sol. Use Sol with max reasoning for plan audits. Never invoke or fall back to GPT-5.5, GPT-5.4, GPT-5.3-Codex-Spark, OSS, or any model older than GPT-5.6.
基于 SOC 职业分类
