Start with one decision.
Complete the Direction Brief, choose the closest playbook, and add System Integrity only when consequence requires it.
Choose one consequential workflow. Define it, trace it, evaluate it, prove its readiness, and record the decision to release—or stop.
Each tool works in your browser without an account. Save explicitly on your device, copy or download Markdown, then move the result into the source of truth your team already governs.
Define the user, decision, evidence, AI role, first-release boundary, evaluation plan, and accountable next action.
Output: one canonical product direction → 02 · DirectTrace one scenario through every product and system layer, its contracts, owners, risks, tests, and next gate.
Output: one inspectable vertical feature row → 03 · Verify AI behaviorMake the dataset, baseline, rubric, thresholds, unacceptable failures, versions, and go/no-go rule explicit.
Output: a repeatable evaluation agreement → 04 · Prove readinessAdvance a feature only when named evidence supports its next state. Keep blockers and residual risk visible.
Output: an evidence-backed readiness state → 05 · DecideRecord scope, approvals, remaining risk, rollout limits, monitoring, rollback, activation authority, and the decision.
Output: go, conditional go, hold, or stop → Completed exampleSee how one Vivid Resume recommendation workflow connects decisions, layers, evaluation, evidence, and release state.
Inspect the connected example →The tools are useful without Gabriel. If the core decision is still unclear, the same artifacts become the working surface for a Direction Sprint.
Complete the Direction Brief, choose the closest playbook, and add System Integrity only when consequence requires it.
Use a five-day Direction Sprint when the user, release boundary, or evidence bar still needs senior product judgment.