Root cause: Use when a Go bug, flaky test, build failure, hang, deadlock, timeout incident, or production regression needs causal diagnosis. Own the first broken invariant and, only when authorized, the smallest causal fix and replayed proof; Skip when feature implementation, unresolved policy design, broad review, or claim-only verification is primary.
Pressure-test: Use when existing research or framing needs a few independent, discriminating questions before specification or planning. Own hidden assumptions, ownership, failure semantics, irreversible consequences, and rollout risk, plus the next owner; Skip when framing is blank, ordinary spec review is needed, or one isolated high-impact approval question is primary.
Approval question: Use when one candidate-spec decision is high-impact, hard to reverse, or explicitly challenged. Own a read-only pressure test with evidence, options, a recommended option, and decision owner; Skip when framing is blank, whole-spec review is needed, or the clarification is low-impact.
Problem frame: Use when a chosen or mostly chosen feature/refactor direction still needs an engineering-ready behavior delta, scope, constraints, assumptions, and readiness decision. Own that frame, its decided defaults, surviving user-owned questions, challenge routing, and next owner; Skip when raw ideation, final design decisions, or task breakdown is primary.
Delivery platform: Use when CI/CD gates, release trust, drift controls, containers, migrations, or rollout policy needs a decision, or when changed delivery controls need conformance review. Own delivery acceptance and review; Skip when API, data, security, reliability, or pipeline implementation is primary.
Test strategy: Use when accepted behavior needs risk scenarios, proof levels, deterministic oracles, or executable gates, or when changed tests and validation evidence need review. Own proof design and test conformance; Skip when behavior is unresolved, tests must be implemented, or completion claims need validation.
Claim-scoped proof: Use when a correctness, readiness, or completion claim needs current evidence of equal scope. Own claim-to-command matching, result inspection, proof scope, and explicit gaps; Skip when code or tests must change, root cause is unknown, test strategy is unresolved, or implementation is still underway.
Performance: Use when latency, throughput, allocation, contention, memory, capacity, or workload budgets need a decision; when a design introduces workload amplification such as per-item I/O or serialization, growing scans, fan-out, batching, retention, or contention; or when changed hot paths and benchmark/profile evidence need review. Own measurable performance policy and conformance; Skip when correctness, reliability, or DB/cache policy is primary.