validate-product

Validate Product Skill

[!IMPORTANT] Validate critical product assumptions and hypotheses through bounded experiment design, precommitted thresholds, and host authorization gates.

Optional args: slug=, ticket=<id/url>, mode=interactive|autonomous|channel, channel=, auto_continue=true|false, profile=business|hybrid|technical.

Instructions

When the user asks to perform this workflow, execute the following steps:

Validate Product Workflow

Goal: Validate prioritized product hypotheses through bounded experiments and precommitted thresholds, routing validated, invalidated, or inconclusive outcomes honestly.

Steps

1. Preflight and intake:
   • Preflight: verify product-management/product-evidence, product-management/product-investment, product-management/product-experimentation; conditional product-management/product-metrics only when designing executable measurement. Scripts: scripts/validate-product-evidence.mjs (ordinary <root>/product-management/product-evidence, Kiro <root>/product-management-product-evidence), scripts/evaluate-investment.mjs (ordinary <root>/product-management/product-investment, Kiro <root>/product-management-product-investment). Validator invocation: node <validator> <ledger.json> --authority-root <canonical-root> --as-of <UTC> --gate experiment (and commit before delivery handoff). Calculator invocation: node <calculator> <scenario.json>. Exit: 0 = coherent (structural, readiness not_assessed) or gate eligible for host check only; never authorization; 3 = proposal_only/gap; 1/2/4 = error (never validated). Emit exact gap strings (<pkg> not installed; opt in at a reviewed ref and sync, product-management/product-evidence excluded; opt in at a reviewed ref and sync, validate-product-evidence.mjs missing; repair package, evaluate-investment.mjs missing; repair package, Node >=20 unavailable; calculation not performed) with safe next action; preserve delivery feature_status. See product-evidence Verification Workflow.
   • Intake: slug, scope, critical HYP records, and strategy/learning authority. Qualitative investment is valid; financial script absence doesn't block qualitative analysis.
2. Design bounded experiment:
   • Define falsifiable EXP: method, target cohort eligibility, sample plan, primary decision metric (MET), guardrails (MET), and precommitted rules (rules: { success, failure, inconclusive }). Never move goalposts after observing data.
   • Predeclare finite bounds: starts_at, ends_at, budget_cap (unsigned decimal string; null cap forbids execution / proposal-only), currency, exposure_cap (safe integer), and stop_rules (host-enforced).
3. Check authorization and run boundary:
   • Verify experiment authorization: proposal in ledger is NOT host approval. Real live user exposure, spend, or feature activation requires authenticated host/operator control; without it, mark proposal_only (AUTHORIZATION_REQUIRED). Never fabricate authorization or execute spend.
   • Material changes to predictions, queries, guardrails, or budget require a new immutable authorized DEC revision, never mutating an existing recorded decision.
4. Evaluate results and route:
   • Completed experiments require observed SRC results and limitations. Inspect negative, confounded, and unseasoned evidence.
   • Route outcomes: proposal-only -> recommended next workflow is null with decision_needed requesting host authorization; invalidated -> route to discover-product (pivot) or retro-learn, or null for defer/no-build (never unconditional implementation); inconclusive -> validate-product (bounded follow-up) or null (defer); supported -> route to plan-feature or implementation-readiness ONLY after recorded, current, scoped investment authorize decision with commit-gate eligibility (eligible_for_host_check, which still requires host check).