Help — PRD Evaluator

How the tool works, how to interpret its output, and what's in scope.

What is the PRD Evaluator?

A PM accelerator. Upload a PRD draft and get structured feedback before formal Step 3 submission. The tool does not approve or reject — it surfaces gaps so the PM can strengthen the document themselves.

Quality vs Estimation-Readiness

Two framings of the same critique:

Quality — Layer 1 substance (problem, hypothesis, outcomes, segment).
Estimation-Readiness — systems, NFRs, dependencies, Cost & Effort Indication.

Layer 1 vs Layer 2

This tool checks Layer 1 (problem definition, JTBD, hypothesis, outcomes). Layer 2 (the solution spec) is intentionally out of scope — Layer 1 must be locked first.

How critique is generated

Workers AI guardrail scans content for sensitive material before it leaves the tenant.
If clear, Claude generates the critique against the v2.4 Step 2 bar.
Both calls are logged. Chat follow-up reuses PRD + critique as context.

Data handling

Your PRD content stays in Novcom's Cloudflare tenant. The Anthropic call goes via [direct API / gateway TBD]. No data is retained beyond the current critique session in the MVP — registry/persistence comes in a later phase.