- Home
- Skills
- Prompt & Skill Engineering
- Skill Compatibility Note Generator
Works with the AI Agents you already use
Skill Compatibility Note Generator
Scans a Claude skill and writes the compatibility note its README needs: surfaces, runtime, packages, tools, network, model and fallbacks, every line backed by evidence from the files
$5
/moSubscription · cancel anytime
Secure checkout via Stripe
- Always the latest version
- Delivered live to your agent via MCP
- Cancel anytime, access ends at period end
Skill Compatibility Note Generator
Example session with this skill installed
Write a compatibility note for my skill in ./youtube-video-analyzer - I want to put it in the README so people know where it runs.
- Read your context and instructions
- Compiled the compatibility note generator
Here's the compatibility note for youtube-video-analyzer. I built it from what the skill actually does: I read SKILL.md and both scripts and ran an evidence scan over the folder. It passes the note validator with no errors, and a copy is saved as COMPATIBILITY.md next to the skill folder.
Compatibility
| Axis | Requirement |
|---|---|
| Surfaces | Claude Code ✅ · Cowork ✅ · claude.ai ⚠️ youtube.com is outside the default package-manager allowlist, so egress must be widened to youtube.com or all domains · Chrome ❌ no shell to run the fetch scripts · API ⚠️ needs a wired-up shell tool with youtube.com egress |
| Runtime | Python 3 — #!/usr/bin/env python3 in both scripts; no version floor declared. Commands call python3 |
| Packages | pip install youtube-transcript-api yt-dlp (add --break-system-packages in a managed sandbox, as SKILL.md:37 does) |
| Tools | Bash (runs scripts/fetch_video.py and scripts/chunk_transcript.py) · file read/write (transcript JSON and chunk files) |
| Network | pypi.org at install time; youtube.com at runtime — both yt-dlp and youtube-transcript-api fetch from it |
| Model | Any model; no vision needed (transcript only). Videos over 2 hours are read across many chunks with running notes — a stronger model keeps them straighter |
| Degrades to | Without yt-dlp: transcript still arrives via youtube-transcript-api, but no title, channel, duration or chapters, so chunks ignore chapter boundaries. Without youtube-transcript-api: yt-dlp subtitles are the fallback, always labelled auto-generated. Without both, or without youtube.com access: the skill stops and says so rather than summarising from the URL |
<sub>Generated from SKILL.md + scripts/ on 2026-10-21. claude.ai network defaults checked against support.claude.com the same day. No items are marked (unverified).</sub>
One-line frontmatter version (if you also want it in SKILL.md; it contains no colon-space, so YAML won't break):
compatibility: Requires Bash and Python 3 with youtube-transcript-api and yt-dlp — needs network access to youtube.com — Claude Code and Cowork, claude.ai only with egress widened
Badge line (for the top of the README):
`Claude Code` · `Cowork` · Python 3 · Bash · youtube.com access
Where each line comes from
- Taken straight from your files:
- The Claude Code / Cowork claim and the youtube.com dependency are in your own frontmatter description.
- The install command is SKILL.md:37.
- The scripts hard-code
https://www.youtube.com/watch?v=...(fetch_video.py:300). - The "Degrades to" row is how
fetch_video.pybehaves. It tries youtube-transcript-api first and falls back to yt-dlp subtitles. Only yt-dlp supplies metadata. If both fail, the script exits withok: false.
Checked against current docs: claude.ai. Its code sandbox allows only package managers (PyPI, npm, GitHub and so on) unless network egress is widened. On Team and Enterprise an owner can add specific domains or allow all domains. So the install step succeeds there, but fetching from youtube.com fails until someone allows that domain.
- Worked out rather than declared:
- The Chrome and API verdicts follow from the skill needing a shell plus youtube.com access.
- The model line is a recommendation, not a requirement.
- "Python 3" has no version floor on purpose: nothing in the skill pins one, so I didn't make one up.
Two things worth fixing in the skill itself
Step 1 installs only when both packages are missing ("If both are missing, install"). If just one is missing, nothing gets installed, and the skill runs in the reduced mode above without saying so. Changing that to "if either is missing" would fix it.
2.
Every command calls python3. On Windows that name is often missing or opens the Microsoft Store, so Windows users of Claude Code may need python instead. If you have Windows users, mention it in the README or make the commands fall back to python.
Do you want me to put the block into the README, or under the first heading of SKILL.md, and add the compatibility: line to the frontmatter?
Subscription · always the latest version
- Secure checkout via Stripe
- Cancel anytime
- Delivered live via MCP
About this skill
Writes the "where does this run" block a skill's README should have, from the skill's own files rather than from guesswork. It runs a scanner over the skill folder for scripts, imports, shell commands, tool calls, URLs and file paths, then classifies what it finds into eight axes and checks the result against a rule set for each surface.
What it does
You get a compatibility table, one line per axis: surfaces (Claude Code, Cowork, claude.ai, Chrome, the API) with a tick, a warning or a cross and the reason for each; runtime; packages with the install command; required tools; network access; model needs; companion skills; and what still works when a dependency is missing. It also gives a one-line frontmatter entry and a badge line, validates the note mechanically, and points out problems it found along the way, such as a command that fails on Windows. Every claim cites the file or line it came from.
Try it first
Point it at a skill folder and say "write a compatibility note for this skill for the README". Check that each line names the evidence behind it.
What it needs
Python 3 to run the bundled scanner and validator, and read access to the skill folder. No network access and no API keys.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Prompt & Skill Engineering.
It reports what the files show, so a requirement the skill never mentions (a binary on the user's PATH, an account) can be missed; it lists what it couldn't settle instead of guessing. It doesn't run the target skill. Surface rules come from a bundled matrix that reflects how the surfaces behaved when it was written, so check edge cases against current docs. It writes the note; inserting it into your files waits for your go.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 27 days ago
- Software
Python
Node.jsSome MCP
Permissions
Browser and Network Access are declared because the skill names curl, Playwright and Chromium: it looks for them in the skill it is checking. It runs a local scanner and makes no network requests and opens no browser itself.