Skip to navigation

Assess Documentation Readiness in DocsFab

Dev
View as Markdown

Run a source-readiness or release-documentation assessment and inspect its scope-bound disposition.

Assess Documentation Readiness

Use readiness to assess whether agents can use the selected repository as input to documentation work. Use release-documentation-readiness to assess public documentation against the governed release rubric. Both jobs are read-only:

docsfab run readiness --repo /absolute/path/to/repository --revision <commit>
docsfab run release-documentation-readiness --repo /absolute/path/to/repository --revision <commit>

Source readiness combines deterministic artifact/inventory observations with explicit model judgments of source completeness and usability. The host checks each judgment’s run, revision, search, and context bindings. A judgment cites the tracked file lines the model read, and the host copies their exact text into the record, so no quote depends on the model reproducing it. A file with credential-shaped content cannot be read for a citation. Missing or unsupported evidence remains specific and visible; an observed required artifact meets the presence criterion. A ready result is an assessment of the declared scope, not permission to publish or release. Rubric judgments belong to the model, so the host does not report them as blocked local probes. If every criterion is unassessed while source material is available, or the model made no source reads, DocsFab asks for another bounded attempt with the source brief and evidence guidance. The CLI and Slack result show the readiness disposition separately from execution status, including unresolved criteria. Slack identifies the retained assessment record; unavailable evidence never implies readiness.

Release readiness requires a valid .docsfab.json committed at the selected revision. Its declared default branch binds the assessment. The run record preserves findings, incomplete evidence, unknown human decisions, and a non-mutating work plan. A successful assessment can still report work required or insufficient evidence. It does not approve a release or authorize publication.

Hosted installations use the same immutable tracked repository evidence and structured outcomes. Operators must configure the workflow and its model composition for the project before its authorized Slack callers can run it.