project-reader-value

Score each project on audience fit, quick-start value, outcome clarity, reusability, novelty, and evidence quality using only provided metadata. Combine these with installation friction and dependency penalties to rank candidates for coverage.

It needs repository metadata, project type, usage scenario, direct purpose, source and hot-list event information, plus the caller's account context. It does not fabricate README details, install steps, or performance claims not present in the input.

No. The evaluation explicitly avoids boosting scores based on star counts, repository age, or trendy names. Every dimension must be justified by the account context and the evidence supplied in the input.

Dimensions that cannot be judged receive a score of 0, and the limitations field documents the evidence gap. This keeps the output honest rather than guessing at unverifiable qualities.

No. The judgments only support candidate ranking in the discovery stage. They do not replace deterministic scoring pipelines or human fact-checking of sources, licenses, and claims before publication.