ToolCairnThe signals
Health blends four families of public signals. None dominates; each one is a partial vote in the composite. The exact weighting evolves as we learn from outcome reports — we treat the formula as a tuning surface, not a public contract.
Reference
Health scoring
Every tool ToolCairn knows about carries a health score between 0 and 100. The number is fine for sorting; the tier is the part that should drive a decision.
Tiers, not just numbers
Active
Maintained, recently shipping, healthy community. Default-pick territory.
Stable
Mature, slower-moving, still alive. Fine to adopt; expect smaller, less frequent updates.
Slowing
Activity has trailed off. Worth a second look at alternatives before committing.
At risk
Inactive, abandoned, or unmaintained. Adopt only with eyes wide open.
Maintenance
How active the project is right now — recency of commits, release cadence, time-to-respond on issues. A tool that has not seen a commit in eighteen months scores low here even with a million stars.
Popularity
Adoption indicators — stars, downloads, dependents. Useful as a tie-breaker but easy to fake; we cap how much popularity alone can move the score.
Activity
Open vs. closed PR ratios, contributor diversity, commit velocity. Catches tools that are popular but stalled.
Community
Documentation completeness, code of conduct, security policy, supported channels. Boring on its own; meaningful when something else fails.
How to read it
Health scores are designed to filter the bad, not crown the best. They eliminate the obvious wrong answers — abandoned packages, drive-by experiments, single-commit repos — so your agent and your team can spend time deciding among genuinely viable options.
Signals versus quality
Health says "is this thing being looked after?" — not "is this code good?". A well-tested, secure, beautifully-architected library that was finished three years ago and never touched again will tier as slowing here. That is the right answer for the question this score is asking; it is not a verdict on the code.
For deeper evaluation — security posture, API ergonomics, performance — pair the health tier with a manual review or pull a comparison via compare_tools.