Understanding trust scores
What our trust score means, how it is calculated, and how to use it in your decisions.
Every listing in the ClawCurrent directory has a trust score from 0 to 100. This number is not a popularity rating or a review average — it's a structured assessment of how safe and reliable a tool is to depend on. Here's exactly how it works.
What the score measures
The trust score is built from five signals, each weighted differently:
Official source verification (30%) — Does the listing link to the tool's real, canonical source? Not a mirror, not a fork, not an unaffiliated repost. We verify this manually.
Maintainer activity (25%) — Is the tool actively maintained? We look at update history, issue response time, and direct maintainer confirmation where possible.
Documentation quality (20%) — Is there clear, accurate documentation? Tools that can't explain themselves clearly tend to behave unpredictably.
Community health (15%) — How active and constructive is the community around this tool? Not just star counts — actual signal quality.
Security review (10%) — For high-trust listings, we run a basic security analysis. Known vulnerabilities are flagged as advisories regardless of score.
What the score ranges mean
90–100: Fully verified. Official source confirmed, actively maintained, comprehensive docs, healthy community, clean security review. Safe to use in production.
75–89: Largely verified. Strong on most signals, possibly missing one (thinner docs, newer project, community still growing). Good choice with minor caveats.
60–74: Partially verified. Some signals are strong, others are weak or missing. Read the breakdown before installing. May be fine for low-stakes use.
Below 60: Review pending or low signal. Either newly listed and not fully reviewed, or has genuine concerns in one or more areas. Treat with caution.
Scores change over time
Trust scores are not static. A tool that was fully maintained can become abandoned. A new project can earn its score over time. We re-review listings on a rolling basis, and maintainers can request re-review after making improvements.
If a tool you rely on drops significantly in score, you'll see it flagged in your dashboard.
When to override the score
Scores are a starting point, not a verdict. A score of 68 on a tool built by a well-known team in a niche domain might be completely trustworthy — the score is low because it's new, not because anything is wrong.
Always read the breakdown, not just the number. The breakdown tells you exactly which signals are strong and which are missing.
A high trust score means we've verified what we can verify. A low score means we haven't — or found something worth flagging. Always read the breakdown before the number.