Investigation Review Hub JSON
Objective
Produce a single JSON object that conforms exactly to the task's local answer template by using the running Investigation Review Hub as the source of record.
Workflow
- Read the prompt and every task-local payload. Extract the matter ID, base URL placeholder or provided base URL, optional SQL/API key header, required top-level keys, enum choices, field definitions, numeric precision, and ordering rules.
- Do not inspect local environment source files, database files, generated manifests, hidden notes, evaluator code, prior answers, or standard answer files. Use only task-local payloads and hub endpoints.
- Query the hub endpoints relevant to the template: schema, matters, categories, productions, custodian sources, documents/search, privilege log, QC findings, retention events, remediation actions, and the read-only SQL endpoint if the task permits it.
- Build an evidence table keyed by stable hub IDs. Track each record's matter, affected categories, status, source/record type, dates, counts, production impact, privilege attributes, and proposed remediation.
- Read references/review-mapping.md before selecting issues, computing category rollups, metrics, or action plans.
- Fill only the keys required by the answer template. Use the template's exact field names, enums, sort orders, integer/null conventions, and JSON-only output rule.
- Validate the final object against the template manually before responding: no prose, no extra keys unless allowed, all required keys present, sorted arrays, stable IDs unchanged, and metrics matching the selected evidence.
Hub Collection Helper
Use the optional helper when a quick endpoint snapshot is useful:
python3 skill/scripts/collect_review_hub.py \
--base-url "$TASK_ENV_BASE_URL" \
--matter-id "$MATTER_ID" \
--api-key-header "$HEADER_NAME" \
--api-key "$API_KEY" \
--out /tmp/review_hub_snapshot.json
Omit --api-key-header and --api-key when the task does not provide them. The helper retries
standard GET endpoints with and without a matter_id query parameter and records failures instead of
guessing facts.
Output Discipline
- Treat the answer template as authoritative over any general legal-review expectation.
- Prefer hub-provided remediation records for action IDs, owners, ranks, priorities, and due dates.
- Keep every category-code list and record-reference list sorted unless the template states another order.
- Use
0for non-applicable numeric counts andnullonly for fields whose template allows null. - Return exactly one JSON object and no markdown, comments, or explanatory text when the task asks for JSON only.