| name | requirements-clarity |
| description | Validate production-ready product requirement completeness across discovery, User Story, and PRD drafting. Use for broad product ideas, requirements documents, admin or operational systems, customer-facing products, workflow-heavy features, regulated or sensitive data, and any request that should be directly implementable and operable rather than automatically reduced to an MVP. |
Production Requirements Clarity
Produce requirements that describe a complete, operable product outcome. Default to production-ready scope. Use MVP, prototype, proof-of-concept, phased delivery, or deferred core capability only when the user explicitly requests that delivery posture.
Completeness Model
Assess the artifact against these levels:
- PC0 Problem unclear: user, pain, current behavior, or desired outcome is unknown.
- PC1 Product intent clear: actors, outcomes, scope, and product surfaces are understood.
- PC2 Workflow complete: entry points, end-to-end flows, states, rules, exceptions, and completion outcomes are testable.
- PC3 Control complete: permissions, data ownership, integrations, notifications, audit needs, and dependency behavior are defined.
- PC4 Production complete: security/privacy, performance/capacity, availability, recovery, observability, support operations, migration, rollout, and accessibility/localization are addressed where applicable.
- PC5 Validated handoff: contradictions are resolved, blocking decisions are answered, acceptance coverage is traceable, and remaining assumptions are explicitly non-blocking.
Do not declare an artifact ready below PC4 unless the user explicitly requests an MVP or exploratory deliverable.
Adaptive Completeness Dimensions
Derive questions from the user's domain. Do not ask a generic checklist when a dimension is irrelevant, but do record an applicability decision.
- Problem and outcomes: who is affected, current workaround, business impact, why now, measurable result.
- Actors and authority: user types, administrators, approvers, external parties, ownership, segregation of duties, permission boundaries.
- Product structure: channels, entry points, navigation, work queues, detail views, reports, settings, and cross-surface handoffs.
- Domain data: objects, required fields, source of truth, validation, identity, relationships, history, retention, deletion, import/export.
- Workflow and lifecycle: trigger, prerequisites, normal path, alternate path, states, transitions, cancellation, reopening, expiration, archival.
- Business rules: calculations, eligibility, ordering, limits, conflicts, duplicate handling, concurrency, idempotency, effective dates.
- Failure and recovery: validation feedback, partial failure, dependency outage, retry, rollback, compensation, manual recovery, escalation.
- Integrations and communication: upstream/downstream systems, API or file boundaries at product level, notifications, templates, timing, delivery failure.
- Trust and governance: authentication, authorization, privacy, consent, sensitive data, audit trail, compliance, abuse and fraud considerations.
- Quality attributes: expected volume, peak load, response expectations, availability, data freshness, accessibility, localization, compatibility.
- Operations: configuration, feature control, monitoring, alerting, support tooling, reconciliation, reporting, audit review, runbook ownership.
- Adoption and change: existing data migration, cutover, backward compatibility, training, rollout, rollback, success monitoring.
Discovery Loop
Use multiple focused question rounds for broad or incomplete requests:
- Intent round: problem, users, outcome, scope, product shape.
- Domain round: objects, workflows, states, rules, permissions, exceptions.
- Production round: integrations, data lifecycle, security/privacy, scale, recovery, operations, migration, rollout.
- Validation round: contradictions, examples/test data, measurable acceptance, unresolved blocking decisions.
Ask two to five related questions per round. Skip answered dimensions. Use multi-select when choices can coexist and single-select only when choices are mutually exclusive. Inferred domain behavior must be presented as a candidate, never silently promoted to fact.
A complete source document does not bypass discovery. Run a gap audit and ask only the missing blocking questions.
Decision Policy
Ask the user when an answer changes product scope, actor authority, money or legal behavior, irreversible data behavior, external integration contracts, lifecycle transitions, or observable acceptance.
Record a non-blocking assumption when it can be changed later without altering those areas. Do not ask product users to choose frameworks, database schemas, packages, deployment topology, or low-level API design unless already stated as a constraint.
Artifact Gates
Requirements
Require product-wide coverage, not a feature list:
- Complete actors, product surfaces, domain glossary, workflows, lifecycle, rules, data, integrations, quality attributes, operations, rollout, acceptance, risks, and open decisions.
- Mark each production dimension as confirmed, not applicable with reason, assumed, or blocking.
- Do not use "future phase" to hide a capability required for safe day-one operation.
User Story
Each story remains a coherent vertical slice and includes:
- Entry point, trigger, preconditions, actor/permission, data and state change, business rules, observable result.
- Acceptance scenarios for normal behavior plus every relevant validation, permission, empty, conflict/concurrency, dependency failure, retry/recovery, and audit outcome.
- Dependencies and operational impact when the slice affects notifications, reports, reconciliation, configuration, or support.
- Traceability to exact requirement sections and explicit non-goals.
Do not create generic stories such as "manage settings" or technical-layer stories. Split when actors, triggers, outcomes, release independence, or acceptance behavior differ.
PRD
Require an engineering-ready product contract:
- Problem/evidence, users/stakeholders, end-to-end journeys, scope/non-goals, functional requirements, business rules, data and analytics, permissions, integrations, NFRs, security/privacy/compliance, operations/support, migration/rollout/rollback, measurable success, risks/dependencies, and traceability.
- Every functional requirement maps to acceptance scenarios and affected production dimensions.
- Open decisions that change scope, money, authority, data integrity, compliance, or acceptance block approval.
Anti-MVP Rule
Do not describe a product as MVP by default. Do not omit administration, permissions, error recovery, auditability, reporting, configuration, migration, or support merely because they are not part of a happy-path screen.
If the user explicitly requests an MVP, still identify production gaps and label them as accepted exclusions with consequences and a reconsideration trigger.