Evidence before build
A task needs independent pain evidence, a clear manual workaround, a real search intent, a technical feasibility boundary, and a reachable first audience.
How tools earn a place in the directory
DocReviewKit starts with a repeated document review problem, then limits the first version to one input, one processing path, and one result a reviewer can use immediately.
A task needs independent pain evidence, a clear manual workaround, a real search intent, a technical feasibility boundary, and a reachable first audience.
Parser claims are tied to fixtures and explicit schemas. Unknown or unsupported review data is recorded instead of inferred.
New tools are added only when the current tools produce qualified task starts, completed exports, result use, or clear repeat-workflow demand.