group-b-content-authoring

name: group-b-content-authoring description: Create or update Group_B_data Logic/Structure markdown files for BPS/WSS/SCPS using the structured blocks schema (groupb_finding/source_observation/method/open_question), with source traceability, confidence, and phase-1 subset decisions.

Group B Content Authoring

Use this skill when editing:

• tools/templates/pts_report_he/Group_B_data/Logic/*.md
• tools/templates/pts_report_he/Group_B_data/Structure/*.md

Goal

Keep Group_B_data as a reliable storage layer for the Hub's normalized knowledge model.

Required structure

Every file must keep:

• YAML front matter (profile_id, doc_kind, status, updated_at, language, schema_version)
• Required ## sections for its kind
• Structured fenced blocks with valid JSON only:
  • groupb_finding
  • groupb_source_observation
  • groupb_method
  • groupb_open_question

Authoring rules

1. Every groupb_finding must include:

• confidence
• status
• source_ids[]
• derivation_method_ids[]

2. Every groupb_source_observation must include:

• source_id
• what_identified_he
• how_identified_he
• artifact_ref
• confidence

3. Prefer explicit line_refs in observations and evidence refs.

4. Keep Hebrew explanations in UI-facing text fields.

• English only for identifiers, code, paths, APIs, spec terms.

5. Add at least one explicit Phase 1 subset decision per profile (logic or structure).

Anti-patterns (avoid)

• Claim without source_ids
• confidence: high on inferred claim without detailed evidence
• Raw markdown prose that duplicates the structured finding without adding value
• Reusing source IDs that are not in source catalogs

Standard workflow

1. Edit one profile at a time (BPS, then WSS, then SCPS)
2. Run:
   • python tools/check_group_b_hub.py
3. Rebuild:
   • python tools/build_pts_report_bundle.py
4. Smoke QA (see skill pts-hub-smoke-qa)
5. Commit in a focused group