HR Policy Explainer

What it takes in

The inputs you provide on each run, validated before anything executes.

2 tasks, run in order.

Each task passes its result to the next. Here's what each does and what it produces.

1. 01
   Answer the question from the policy only. Read {{policyText}} and answer {{employeeQuestion}} using ONLY what the policy actually says. Quote the relevant clauses verbatim, citing their section names or numbers where the text provides them. Clearly separate what the policy clearly states from what it does not cover or only partially addresses. Never invent policy, fill gaps with assumptions, or generalize from common HR practice; if the policy is silent on part of the question, state that explicitly. Present readably, never JSON.

   Expected output

   A readable, grounded answer with no JSON and no code blocks containing: Answer (a direct response to the question based strictly on the policy), Relevant Clauses (verbatim quotes with section references where available), What The Policy Clearly Says (bullets), and What The Policy Does Not Cover (bullets naming any part of the question the text is silent or only partial on, or a note that the policy fully answers the question).
2. 02
   Wrap up with guidance. Using the answer from Task 1, produce the employee-facing wrap-up: a plain-English summary of the answer in two to four sentences, a list of any ambiguity in the policy that needs a human HR judgment call (with a one-line explanation of why each is ambiguous), and, if the policy did not fully answer {{employeeQuestion}}, a short ready-to-send escalation note the employee can email to HR that states the question, what the policy says, and exactly what clarification is needed. If the policy fully answered the question, say so and skip the escalation note. Present readably, never JSON.

   Expected output

   A readable wrap-up with no JSON and no code blocks containing: Plain-English Summary (2 to 4 sentences), Needs Human Judgment (bullets listing each ambiguity and why it requires HR's call, or a note that nothing is ambiguous), and Suggested Escalation Note (a short ready-to-send email with a Subject line and body, included only if the policy did not fully answer the question).