Listing standard
One template for every connector, skill and plugin listing, written as eight checks a script can run. Every listing is audited against it, the most-viewed broken listings are rewritten first, and every new release must pass all eight before launch.
The template
| Field | Rule |
|---|---|
| Title | Product name only, up to 30 characters. No "Connector", "MCP", "Integration", "Plugin", "Skill", no superlatives, no all caps. |
| One-line summary | One sentence, up to 90 characters, ending in a period. Leads with what the user can do, not with the company's name. |
| Capabilities | 3–5 bullets, each starting with a capitalized verb, up to 60 characters. |
| Permissions | Always stated (e.g. "Read", "Read and write (pages you can already access)"). |
| Required connectors | Always stated; "None" when there are none. |
| "Try asking" | 3+ prompts a user could paste, 15–80 characters each. |
| Author | Always shown. |
| Attribution | Ports name the source repo, the creator's handle and the license: "Ported from owner/repo by @creator (MIT)". |
Where the catalog falls short
How to read this: share of all listings that pass each check today. The shortest bars are where a sweep pays off most.
Rewrites of high-traffic listings
Top-priority listings
How to read this: priority = monthly views × share of checks failed, so a popular, broken listing comes first. The top are rewritten at a week; the rest are fixed in the next catalog sweep.
Every listing
Pre-launch checklist
The same eight checks run on every new release before it ships, so new listings never add to the cleanup. This week: the sketchwright listing and the draft listing in the Brightpath Analytics build brief both pass 8/8. In the GitHub export workflow, a failing check blocks the merge.