doc-traceability-validator

Run validate_features.py with --code-dir pointing at your source tree and --docs-file pointing at features.md. It reports undocumented features, orphaned entries, and duplicate IDs with completeness and uniqueness scores.

Use validate_traceability.py with the --features, --rules, and --usecases arguments. It extracts FEAT, BR, and UC references from each document and reports broken links plus cross-reference coverage percentages.

No external dependencies are required. All scripts use only the Python 3.9+ standard library, including re, pathlib, argparse, json, and dataclasses, so they run in any CI environment without pip installs.

Yes, use the --fail-threshold flag on validate_features.py or --fail-on-broken on validate_traceability.py. The scripts exit with code 1 when scores fall below the threshold or broken references exist, failing the CI job.

The validator distinguishes cross-cutting duplicates from true collisions. The same feature ID spanning multiple architectural layers like types, stores, hooks, and components is intentional in React/TypeScript codebases, while same-layer duplicates are flagged as collisions.

Run generate_registry.py with --code-dir and --output. It scans for @implements annotations, infers descriptions from doc comments, groups entries by module namespace, and writes markdown entries plus an index table.