PeopleOps Console Solver
Solve PeopleOps Console verification and reconciliation tasks by gathering structured data from the task environment REST API, applying the domain's record-precedence rules, and producing a JSON answer that conforms to the supplied answer template.
How to approach a PeopleOps task
Every PeopleOps task follows the same pattern. Treat each step as a gather-then-decide phase; don't jump to conclusions before you have all the evidence.
1. Connect and orient
The task will give you:
- A base URL for the task environment (shown as
<TASK_ENV_BASE_URL>) - Login credentials (always
ops.lead@peopleops.local/PeopleOps#2026in these tasks) - A business goal describing what to verify and which employee or case to inspect
Start by reading /api/summary to confirm the environment is reachable and understand overall counts. Then pull the broad datasets you'll need (employees, cases, payroll ledgers, policies, audit events, documents, messages, recruitment) with the GET endpoints listed in references/api-endpoints.md. Don't read just one record at a time when the full list is small; pull the full collection and filter client-side.
2. Gather targeted detail
Once you know which employee ID, case ID, or requisition ID the task names, drill into the detail endpoints:
- Employee profile:
/api/employees(filter byemployee_id) - Case detail with approvals, attachments, and embedded audit events:
/api/cases/{case_id} - Policy detail:
/api/policies/{policy_id} - Single audit event:
/api/audit/{audit_id} - Recruitment packet:
/api/recruitment(filter byopening_id)
The ledger endpoint (/api/payroll-ledgers) returns leave assignments, salary assignments, payroll worksheets, HRMS leave ledgers, and People Ops adjustments in one flat list. Filter by employee_id and record_type to isolate what you need.
3. Apply business precedence rules
This is the core of every PeopleOps task. Read references/business-rules.md for the full rule set. The essential ones:
| Rule | How to apply |
|---|---|
| Submitted > Draft | A record with status: "Submitted" controls over one with status: "Draft". Draft records must be excluded from authoritative answers. |
| Approved > Superseded | When multiple leave/payroll assignments exist for the same employee and period, the latest status: "Approved" or "Submitted" record controls. Exclude "Superseded" and "Draft" records. |
| Approved assignment > Profile summary | When an approved leave assignment from the ledger conflicts with the employee profile's leave_balance_days or implied policy, the approved assignment controls. Mark the profile as stale/ignored. |
| Folder readiness | A folder is ready only when all required_files and all required_tags are present in the document record. |
| Notice quality | A formal notice is valid only when its defects array is empty. Each defect maps to a specific label from the answer template. |
| Recruitment handoff | Payroll handoff is created only for the selected candidate with an accepted offer. Draft payroll prechecks do not satisfy the gate. |
4. Scope audit evidence correctly
Many tasks ask you to identify supporting audit events and exclude irrelevant ones. Read all audit events for the target employee or case. Include only events whose event field and detail field are directly about the scope described in the task (e.g., leave source precedence, document/notice findings, payroll readiness). Exclude events about unrelated scopes even if they share the same employee or case ID.
5. Map to the answer template
The task supplies an answer_template.json in input/payloads/. Every field in that template has a type constraint and, for enums, an allowed_values list. Your JSON answer must:
- Include every key from the template (no missing fields)
- Use exactly the normalized labels from the template's
allowed_valuesfor every enum field — never invent free-text alternatives - Match the declared types exactly (string, number, integer, boolean, list[string])
- For list fields, return plain arrays of string IDs or enum labels, never objects
Before finalizing, reread each template key and confirm your answer value exists in the allowed_values list for that key. If a template key has no allowed_values (it's a plain string or number), derive the value from the authoritative data you gathered.
6. Validate and return
Cross-check your answer against the raw API data one more time. Specifically verify:
- Every excluded record ID appears in the raw data with a status that justifies exclusion (Draft, Superseded, or stale)
- Every selected record ID appears in the raw data with a status that justifies selection (Approved, Submitted, or the controlling source)
- Sums (like recruitment cost total) are correct arithmetic on the source ledger items
- Boolean fields (folder_ready, required_tag_present, accrual_ready, profile_policy_ignored) are grounded in explicit data, not assumptions
Return only the JSON object. No markdown fences, no explanatory text — unless the task explicitly asks for something else.