Grok Api

通过 LeonAI 代理调用 xAI Grok API，支持多模型文本对话、深度推理及图片视频生成与编辑，兼容 OpenAI 格式。

Security Scan

The skill's code and docs match its stated purpose (calling Grok via a LeonAI proxy) but there are metadata inconsistencies and a third‑party proxy endpoint that you should verify before supplying an API key.

Purpose & Capability

The name/description (call Grok via LeonAI proxy) aligns with the included code and SKILL.md: both show HTTP calls to a Grok API proxy and support chat/image/video endpoints. However the registry metadata lists no required environment variables while SKILL.md and the script expect GROK_API_KEY and GROK_BASE_URL, which is inconsistent and could lead to user confusion.

Instruction Scope

Instructions only direct the agent/user to call the proxy API endpoints (chat, images, videos) and to store GROK_API_KEY in personal-secrets.json; they do not instruct reading unrelated files or accessing other credentials. They do instruct sending potentially sensitive content to the configured BASE_URL, so the trustworthiness of that endpoint matters.

Install Mechanism

No install spec or remote downloads; the skill is instruction-only with a small pure-Python script that uses only the standard library. Nothing is written to disk beyond the included script when you install the skill.

Credentials

Only an API key and a base URL are required by the script, which is proportionate to the task. The concern is twofold: (1) the registry metadata does not declare these required env vars, and (2) the default BASE_URL points to a third-party proxy (https://apileon.leonai.top). Supplying an API key to an unverified proxy can expose your requests and data to that operator.

Persistence & Privilege

The skill does not request always:true, does not modify other skills or system config, and has no persistent/install-time privileged behavior. Agent autonomy remains at the platform default.

What to consider before installing

This skill appears to do what it says (calls Grok via a LeonAI proxy) and the Python script is small and readable. Before installing or providing keys: (1) verify who operates the default proxy URL (apileon.leonai.top) — if you don't trust it, change GROK_BASE_URL to an endpoint you trust or use an account/key dedicated to the proxy; (2) do not reuse high-privilege keys — create a scoped/test key if possible and avoid sending secrets or sensitive data through the proxy; (3) note the registry metadata omission: the skill requires GROK_API_KEY and GROK_BASE_URL even though the registry lists none, so make sure your environment/set-up matches the SKILL.md; (4) review the small grok_chat.py yourself (it only does HTTP calls) and test with non-sensitive inputs first. If you cannot verify the proxy operator, consider using an official xAI endpoint or another vetted integration instead.

Grok API Skill

Description

通过 LeonAI 代理调用 xAI Grok API，支持文本对话、推理、图片生成/编辑、视频生成。兼容 OpenAI Chat Completions 格式。

触发场景

• 用户说"用 Grok 聊天/问问题/分析"
• 用户说"用 Grok 生成图片/画图"
• 用户说"用 Grok 生成视频"
• 用户提到 Grok 模型名（grok-3, grok-4, grok-4.1 等）

环境变量要求

• GROK_API_KEY — LeonAI 代理的 API Key（存入 personal-secrets.json）
• GROK_BASE_URL — 默认 https://apileon.leonai.top/grok/v1

可用模型

文本模型

模型	特点	Tier	Cost
grok-3	基础文本	basic	low
grok-3-mini	轻量推理	basic	low
grok-3-thinking	深度推理	basic	low
grok-4	进阶文本	basic	low
grok-4-mini	快速回复	basic	low
grok-4-thinking	进阶推理	basic	low
grok-4-heavy	重量级	super	high
grok-4.1-mini	最新快速	basic	low
grok-4.1-fast	极速	basic	low
grok-4.1-expert	专家级	basic	high
grok-4.1-thinking	最新推理	basic	high
grok-4.20-beta	Beta测试	basic	low

图片/视频模型

模型	功能
grok-imagine-1.0	图片生成
grok-imagine-1.0-fast	快速图片生成
grok-imagine-1.0-edit	图片编辑
grok-imagine-1.0-video	视频生成

使用方式

1. 文本对话

curl -s -X POST "$GROK_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $GROK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-4.1-mini",
    "stream": false,
    "messages": [{"role": "user", "content": "你好"}]
  }'

2. 推理模式

curl -s -X POST "$GROK_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $GROK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-4.1-thinking",
    "stream": false,
    "reasoning_effort": "high",
    "messages": [{"role": "user", "content": "分析量子计算的前景"}]
  }'

3. 图片生成

curl -s -X POST "$GROK_BASE_URL/images/generations" \
  -H "Authorization: Bearer $GROK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "描述文本",
    "model": "grok-imagine-1.0",
    "n": 1,
    "size": "1024x1024",
    "response_format": "url",
    "enable_nsfw": true
  }'

可选 size: 1024x1024 / 1280x720 / 720x1280 / 1792x1024 / 1024x1792

4. 视频生成

curl -s -X POST "$GROK_BASE_URL/videos" \
  -H "Authorization: Bearer $GROK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "视频描述",
    "model": "grok-imagine-1.0-video",
    "size": "1792x1024",
    "seconds": 6,
    "quality": "standard"
  }'

seconds: 6-30 | quality: standard(480p) / high(720p)

5. 脚本调用

# 文本
python3 scripts/grok_chat.py "你好Grok"
python3 scripts/grok_chat.py --model grok-4.1-thinking --reasoning high "分析问题"

# 图片
python3 scripts/grok_chat.py --image "描述"

# 视频
python3 scripts/grok_chat.py --video "描述"

# 列出模型
python3 scripts/grok_chat.py --list-models

推荐模型选择

• 日常对话/快速问答 → grok-4.1-mini 或 grok-4.1-fast
• 深度分析/推理 → grok-4.1-thinking + reasoning_effort: high
• 图片生成 → grok-imagine-1.0（高质量）或 grok-imagine-1.0-fast（快速）
• 视频生成 → grok-imagine-1.0-video

注意事项

• API 通过 LeonAI 代理转发，非 xAI 直连
• 额度：Basic tier 80次/20h，Super tier 140次/2h
• 图片/视频生成为异步，URL 有时效性，建议及时下载
• enable_nsfw: true 可开启 NSFW 内容生成