How we rate & verify.

How genuinely useful a tool is — independent of its GitHub stars. A hello-world demo inside a famous repository stays at one flame.

• 🔥🔥🔥🔥🔥 — transformative and broadly useful; foundational infrastructure or best-in-class at a major task
• 🔥🔥🔥🔥 — very useful with solid scope
• 🔥🔥🔥 — useful in its niche
• 🔥🔥 — narrow, duplicative, or only meaningful inside one project's own workflow
• 🔥 — trivial, demo, or template

Each rating comes with a one-sentence explanation shown in the tool's Quick Facts, so you can see why it earned its flames.

A tool is Verified when an AI review of its documentation confirms all four of the following:

1. Documentation exists and has real substance
2. The claims match the visible mechanics — concrete tool lists, commands, and configuration back up what the description promises
3. No overclaiming beyond what the docs demonstrate
4. The pieces are coherent (e.g. required credentials match the claimed integrations)

What Verified does not mean (yet): we have not executed the tool. Runtime verification — actually installing and invoking top-rated tools in a sandbox — is on our roadmap and will be labelled separately when it ships.

• Free — open source; no account, no credit card
• Account required — needs an API key or account; the provider may offer a free tier
• Paid service — backed by a paid product; expect usage costs

Every tool's Quick Facts also lists the exact credentials it needs and its open-source license, so you are never surprised by a paywall or a permission ask.

Catalogs rot: repositories get renamed, skills move to new folders, projects disappear. Every week we re-check each tool against GitHub itself:

• Renamed or transferred repositories → links updated automatically
• Skills relocated inside their repository → deep links re-pointed
• Confirmed-gone projects → removed from listings (double-checked first — a moved project is never mistaken for a dead one)
• Documentation and licenses re-synced from the source