How the TrustScore works

One transparent formula, applied identically to every repo on every run. We publish the weights so you can argue with them — that is the point.

Base score
100
Every repo starts organic.
Evidence
GH Archive
10B+ events since 2011, refreshed hourly.
Repo metadata
GitHub API
Stars, forks, age, owner type.
SignalTriggerPenaltyWhy
Star-farm overlap>15% of starrers starred 3+ repos in the windowup to −60Rings rent accounts that churn through trending repos.
Single-hour dump>50% of day's stars in the busiest hourup to −25Organic attention spreads across time zones.
No dev engagement0 issues/PRs/pushes & ≤1 fork at ≥10 stars (−25; orgs −10)−25 / −10Star-only attention is the purchased-growth signature.
Low dev engagement<3 dev events at ≥10 stars (orgs −5)−12 / −5Same, softer case.
Lifetime spike>90% of all-time stars arrived today−20New repos only — one-day history is not organic history.
Dormant starrers>70% starrers inactive for 90dup to −12Weak signal: lurkers are normal. Kept small deliberately.
Spam keywordsknown spam/malware terms in name−20Cheat/crack/keygen families. Shown, never silent.
Farm usernamebatch-registration name pattern−15Auto-generated handles of farmed accounts.
Grades: ≥80 Verified Organic · 60–79 Mixed Signals · 40–59 Suspicious · <40 Flagged.
Known limits: heuristics, not proof — a report explains evidence, you make the call. Campaigns that spread over days with aged accounts can pass; that is why we publish every penalty. Dispute a report any time — we adjust weights publicly.

Generated for board 2026-11-22.