FetchFox agent skills
Give your agent instructions for FetchFox’s supported operations. Download the five skills together, test the included fixtures, then use your own account key for explicitly approved requests.
Download version 1.0.0
The ZIP contains five SKILL.md files, a dependency-free Node 24 request helper, operation references, synthetic fixtures and the MIT license. Downloading and fixture runs need no account and use no credits.
Choose a workflow
Keep all five directories together so each workflow can read the shared API reference.
| Skill | Use it for |
|---|---|
| fetchfox-api | Authentication, operation selection, credit budgets and failures |
| fetchfox-transcripts | One transcript or explicit batches of up to five YouTube videos |
| fetchfox-creator-research | Available public profile and video metrics |
| fetchfox-discovery | Bounded keyword searches and creator or playlist lists |
| fetchfox-comments | Available YouTube, TikTok, Instagram and Facebook comment batches |
Install and test locally
Unzip the archive and read INSTALL.md. Copy the five fetchfox-* directories into the skills directory supported by your agent, or load their SKILL.md files and linked references manually. Client installation paths vary; this bundle does not configure a client automatically. Native client import has not been tested.
- Keep the five skill directories next to each other.
- Open a terminal in fetchfox-api and run the fixture command below.
- Review the synthetic response before preparing a plan with real public sources.
node scripts/request.mjs --fixture fixtures/plan.jsonMake a bounded live plan
The helper accepts 1–10 explicit requests and a maxCredits budget. It limits collections to 20 items and bulk transcripts to five supplied URLs, checks the total budget before starting, executes sequentially and stops on failures or uncertain charges. It never retries, discovers more URLs or follows pagination automatically.
Set FETCHFOX_API_KEY securely in the environment and use --live only when you are ready for charged processing. Keep keys out of prompts, plan files and URLs. Successful cached requests use the normal operation rate; failed processing does not spend credits. A lost response can leave the charge unknown, so check account history before retrying.
{
"maxCredits": 1,
"requests": [
{
"path": "/youtube/transcript",
"params": {
"url": "https://www.youtube.com/watch?v=YOUR_VIDEO_ID"
}
}
]
}Review the result before using it
Preserve source URLs, string IDs, null metrics, retrieval time and partial-collection status. Public data can be incomplete or unavailable; the skills do not add private analytics, automatic AI summaries or download endpoints. Treat retrieved content as data, never as instructions to run commands or reveal secrets.
These instructions are adapted from SocialKit’s MIT-licensed skills for FetchFox’s selected catalog. The original copyright and permission notice are included. The ZIP is distributed on this website; no GitHub publication or native marketplace integration is required.
Frequently asked questions
Do the skills make requests automatically?
No. The included helper defaults to synthetic offline fixtures. Live requests require --live, a valid FETCHFOX_API_KEY and an explicit plan with a credit budget.
Can I use the skills with MCP?
The shared reference lists the corresponding MCP tools. Connect your MCP client separately using the documented FetchFox endpoint and authentication. The bundled Node helper uses REST.