rai-discover-validate

Automate discovery analysis validation using a confidence-tier workflow that auto-validates high-confidence components, batch-reviews medium-confidence modules, and flags low-confidence items for individual human review. This reduces manual checks from individual components to module-level batching.

Confidence-tier validation is a process that categorizes analysis artifacts by score, auto-validating entries with confidence >= 70, batching medium-confidence items by module to synthesize purposes and categories, and isolating low-confidence items for detailed manual review.

Finalize a components-draft by running the validation workflow to auto-populate auto_purpose, auto_category, and validation flags, saving updates to work/discovery/components-draft.yaml and exporting the validated data to components-validated.json.

Discovery validation preserves audit trails throughout the per-module batch processing workflow. It maintains validation flags and detailed signals for low-confidence components while preparing the validated data for subsequent graph integration.

You need to run a discovery scan first to produce analysis artifacts. The validation workflow applies to outputs where a scan generated the initial data, requiring these artifacts to begin auto-validating and batch-reviewing the components-draft.