quorum-degradationlisted

# Degradation The gate never gets skipped for lack of capability. It **degrades**, and it says which tier it ran at. This matters because a managed or restricted seat may block the Workflow tool, subagents, or both — and the temptation is to skip the review entirely rather than run a weaker one. A weaker gate honestly labelled beats no gate. > **Convention, not measurement.** The three tiers are a portability design. The corpus measured the > Tier 1 shape; Tiers 2 and 3 are reasoned degradations of it. See [CALIBRATION.md](../../CALIBRATION.md). --- ## Tier 1 — parallel subagents The full shape. Miners run concurrently and blind to each other; verifiers run per finding; judges see only the confirmed dossier. **What makes it strongest: genuine independence.** Reviewers who cannot see each other's work do not converge, and non-convergence is what produces coverage. Requires a workflow/orchestration tool. Use `workflows/quorum.js` or scaffold one with `tools/scaffold.mjs`. --- ## Tier 2 — sequential subagents Subagents available, no parallelism. Same lenses, run one at a time. **The rule that preserves most of the value: each pass gets a fresh context and does not see earlier findings.** Independence, not concurrency, is what Tier 1 was buying. A sequential run that hides prior results retains most of the rigour; one that accumulates them collapses into a single perspective that agrees with itself. Costs versus Tier 1: wall-clock time, and some risk of orchestrator