grok-query-scrape
Submit prompts to https://grok.com using an authenticated cookie file and capture the generated answer verbatim. Use when the task requires raw prompt passthrough, raw Grok output capture, and reproducible artifacts without JSON parsing.
来源信息
- 仓库
- yang0/skill_grok
- 最近来源活动
- 2026年2月28日 08:06
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 0
- Forks
- 0
安装方式
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。
检查来源文件
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。
文件资源管理器
5 个文件正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- grok-query-scrape
- description
- Submit prompts to https://grok.com using an authenticated cookie file and capture the generated answer verbatim. Use when the task requires raw prompt passthrough, raw Grok output capture, and reproducible artifacts without JSON parsing.
# Grok Query Scrape
Use this skill to run a repeatable Grok automation workflow:
- Load Netscape-format cookies (for logged-in session reuse)
- Open Grok in a browser session
- Submit the prompt exactly as provided
- Wait for answer generation to stabilize
- Return and save raw output only (no parsing)
Language guides:
- Chinese guide: [references/usage_zh.md](references/usage_zh.md)
- English guide: [references/usage_en.md](references/usage_en.md)
## Workflow
1. Install runtime dependencies when missing:
```powershell
pip install patchright
```
2. Prepare a cookie file in Netscape format (example: `H:\cookies\grok.txt`).
3. Run the script:
```powershell
python E:\projectHome\skill_grok\skill\grok-query-scrape\scripts\query_grok.py `
--cookie-file H:\cookies\grok.txt `
--prompt "Find top 10 AI productivity tweets from the last 24 hours." `
--output-dir E:\ `
--basename grok_ai_top10 `
--wait-seconds 240
```
4. Check outputs:
- `<stdout>`: Grok response text (verbatim capture from `main`)
- `<basename>_raw.txt`: same raw Grok response text saved to file
- `<basename>_meta.json`: run metadata and paths
## Script
Use [query_grok.py](scripts/query_grok.py) for execution.
Key flags:
- `--cookie-file`: Netscape cookie txt path (required)
- `--prompt`: inline prompt
- `--prompt-file`: prompt text file (alternative to `--prompt`)
- `--output-dir`: output folder (default `.`)
- `--basename`: output filename prefix (default `grok_result`)
- `--wait-seconds`: max wait for response stabilization (default `180`)
- `--headless`: run browser headless
- `--new-chat`: send `Ctrl+J` before query to reduce contamination (enabled by default)
## Prompting Guidance
This skill forwards the prompt as-is, so put all constraints directly into your prompt text.
## Failure Handling
If the output format is not what you need:
1. Keep `_raw.txt` as ground truth.
2. Tighten the prompt and retry.
3. Keep `--new-chat` enabled for isolation.
4. Increase `--wait-seconds` when generation is long.
If login fails:
1. Refresh cookie file from a valid browser session.
2. Keep domain/path/secure flags intact in Netscape format.
3. Retry with `--headless` disabled to inspect behavior.
### scripts/
- [query_grok.py](scripts/query_grok.py): submit prompt, wait for answer, return/save raw/meta only.
在 GitHub 查看