Keep metrics distinct
Repository stars, Skill quality, and test results should not collapse into an opaque total score.
Ranking index
Different questions need different rankings
Trend, quality, freshness, official-source status, and test evidence answer different questions. No single composite score should replace the decision.
- Page type
- Ranking boards
- Publish rule
- Evidence-gated
- Primary job
- Support decisions
Ranking boards
Every board publishes its method and limits
Ranking pages disclose data ownership, refresh dates, cold-start behavior, and what each signal cannot prove.
EvidenceSourceComputedTestedEditorial
ComputedDaily
Trending
Discover recently rising projects using seven-day source-repository attention changes.
ExploreComputedPublic formula
Quality
An explainable quality score built from documentation, specificity, maintenance, and trust signals.
ExploreSourceFreshness
Recently updated
Review Skills whose source files or repositories have recent activity.
ExplorePilotNoindex
Official-source pilot
Separate source ownership from verified official status; remain a pilot until identity checks are complete.
ExploreTestedEvidence-gated
Tested pilot
Accept only reproducible tests with environment, input, result, limits, and date.
ExplorePage admission
Not every combination deserves a URL
Every aggregation page needs a distinct decision job, enough qualified entities, and useful content beyond a repeated card list.
Disclose cold starts
When snapshot history is incomplete, fallback ordering is disclosed.
Meet the ranking gate
A live board needs at least 20 qualified entries, a public method, and a visible update date.
EntityIdentity, source, and lifecycle verified
DemandThe page answers a distinct user question
EvidenceAt least two unique value classes
Continuous learning
Pages evolve with real search and product behavior
GSC queries, internal search, saves, install copies, source clicks, and maintenance cost determine whether a page is enriched, merged, paused, or scaled.