SkillRank Data Methodology

SkillRank tracks vendor model products and GitHub-backed agent or developer-tool repositories. Vendor model profiles are curated because many closed-source models do not expose public repository telemetry.

GitHub-backed entries are verified against the GitHub REST API during the daily crawl. When verification succeeds, the site records stars, forks, language, license, topics, and repository update dates.

Scores are directional comparison aids, not universal truth. They combine category fit, product maturity, integration value, and available freshness signals. A higher score should prompt a closer look, not replace procurement diligence.

Coding and repository-backed entries can move when GitHub activity changes. Closed vendor models require editorial updates when official release notes, pricing pages, or product documentation changes.

The daily crawler records score, rank, and verified GitHub stars into a local history file. Trend charts appear from recorded snapshots rather than fabricated projections.

Fresh installations may show a single current snapshot until several daily runs have accumulated. This is intentional: a short truthful history is better than a long simulated one.

Readers and maintainers can send corrections through the contact page. High-impact corrections should include an official source, repository URL, release note, or pricing page.

SkillRank favors conservative updates: if a claim cannot be verified from an official source or a public repository, it should be framed as editorial context rather than a hard fact.