Add WebHarbor site registry audit script

[Lxr-max]

Summary

• adds scripts/audit_site_registry.py to audit WebHarbor site registration consistency, port mappings, reset integration, task URL wiring, and lightweight asset metadata
• adds scripts/test_audit_site_registry.py with standard-library tests for duplicate ports, missing registrations, missing directories, strict mode, and JSON output
• adds a short README usage section for the new audit tool

Why this helps WebHarbor review

• catches registry drift between websyn_start.sh and control_server.py
• surfaces port collisions and Docker exposure gaps before contributors open site PRs
• checks task files are wired to the registered localhost port without duplicating the full task validator
• gives reviewers a fast repo-level health check for site registration changes

CLI usage

python scripts/audit_site_registry.py
python scripts/audit_site_registry.py --site amazon
python scripts/audit_site_registry.py --strict
python scripts/audit_site_registry.py --json

Checks included

• site discovery across sites/, websyn_start.sh, and control_server.py
• registry list mismatches and duplicate site entries
• port mapping and Docker EXPOSE coverage checks
• reset example consistency from README
• task JSONL parse + task localhost port integration checks
• runtime file hygiene warnings for instance/, scraped_data/, logs, caches, and screenshots
• lightweight .assetpaths coverage checks without fetching or editing assets

Test commands and results

• py -m py_compile scripts/audit_site_registry.py ✅
• py -m py_compile scripts/test_audit_site_registry.py ✅
• py scripts/test_audit_site_registry.py ✅ (7 tests passed)
• py scripts/audit_site_registry.py ✅
• py scripts/audit_site_registry.py --json ✅

Audit run result on current repo

• 15 site directories found
• 15 registered sites found
• 15 registered ports found
• 15 task files checked
• 643 tasks scanned for registry/task integration
• 0 errors, 0 warnings on current upstream/main baseline

Notes

• this is not a website contribution
• no HF assets are involved
• .assets-revision was not modified
• no existing site implementation was changed

Known limitations

• asset checks are metadata-only and intentionally do not require local HF assets to be fetched
• task checks focus on registry and port integration, not the full task-content validation covered by the separate task-validator tooling