By name: bubbles-quality-gates-catalog description: Look up Bubbles quality gates (G024–G100+) by ID, by symptom, or by enforcement script. Use when a guard rejects work and you need to understand the failing gate, when designing a scope's DoD, or when wiring tests to cover gate-relevant scenarios. Covers the canonical test taxonomy, gate IDs, and which script enforces each one.
Bubbles Quality Gates Catalog
Goal
Resolve "what is gate G0XX" and "which script enforces this" without re-reading the entire governance corpus. Use this as the entry point into the gate catalog.
When to use
- A guard (state-transition, artifact-lint, regression-quality, regression-baseline, etc.) failed with a
G0XXlabel - Designing a scope's DoD and need to know which gates apply
- Writing a test and need to know which gate IDs the test should cover
- Auditing whether a scope's DoD covers all gates relevant to the change
v4.1.0 gate refinements
Gates G073, G090, G022, G041, G040, G008A, G009, G056 received non-breaking refinements in v4.1.0. See docs/v4.1.0-delivered-pending-activation.md for the per-gate behavior change and how scopes opt in via Scope-Kind:, Lockdown-FRs:, deliverableFiles[], phaseStubs{}, and report.md anchor references.
High-frequency gate IDs (memorize these)
| Gate | Topic | Enforced by |
|---|---|---|
| G023 | State transition guard envelope | state-transition-guard.sh |
| G024 | Spec cannot be done until all scopes Done |
state-transition-guard.sh, artifact-lint.sh |
| G025 | Per-DoD-item inline raw evidence ≥10 lines | artifact-lint.sh, state-transition-guard.sh |
| G026 | Stress tests required when scope defines latency SLAs | artifact-lint.sh |
| G027 | Phase-Scope coherence (completedPhases ↔ completedScopes) | state-transition-guard.sh |
| G028 | Implementation reality scan (no stubs/fakes/hardcoded data) | implementation-reality-scan.sh |
| G029 | Integration completeness (every artifact has a consumer) | state-transition-guard.sh |
| G035 | Vertical slice completeness (cross-layer scopes) | state-transition-guard.sh |
| G040 | Deferral language detection | state-transition-guard.sh Check 18 |
| G042–G068 | Capability delegation, policy provenance, validate certification, scenario manifest, lockdown, regression contract, scenario TDD, rework packets, owner-only remediation, concrete results, child-workflow depth | workflows.yaml + targeted guards |
| G073 | Mode ceiling pre-flight (statusCeiling enforcement) | state-transition-guard.sh Check 3B |
| G079 | Impact-aware validation (testImpact config) | test-impact-plan.sh |
| G080 | Trace contracts (traceContracts config) | trace-contract-guard.sh |
| G082–G093 | Convergence cap, compaction discipline, pre-existing deferral, dogfood, orchestrator persistence, planning packet linkage, post-cert edits, inter-spec deps, retro convergence, planning chain, strict terminal status, delivery delta | targeted guards (each named after the gate) |
| G094 | Capability foundation gate | capability-foundation-guard.sh |
| G095 | Discovered-issue disposition (every observed issue is filed, not deferred) | discovered-issue-disposition-guard.sh |
| G097 | Requirement-mechanism correspondence (named mechanism ↔ code, warn-and-justify) | requirement-mechanism-guard.sh |
| G098 | Observability posture declared (wired/opted-out; undeclared WARN, project-flippable to block; fake-wired / opted-out-malformed / unsupported-schema fail loud; framework source EXEMPT) | observability-posture-guard.sh |
| G099 | Observability opt-out freshness (recorded + expiring opt-out; missing required optOut field fails loud; expired revisitAfter = non-blocking route-required reminder) | observability-opt-out-guard.sh |
| G100 | Observability SLO evidence (BLOCKING when posture: wired + an instrumented scope's observabilityWorkflow targets a workflow with an slo: link; captured .specify/runtime/observability/<workflow>.slo.json must MEET the traceContracts.observability.slos target; a contract-declared metric absent from observed is a breach; malformed / wrong-workflow evidence fails loud before numeric compare; missing jq/yq fails CLOSED; framework source EXEMPT). Division of labor: G026 ensures the stress/load test EXISTS + cites the SLO registry; G100 ensures the captured evidence MEETS the target. |
observability-slo-guard.sh |
| G101 | Release-delivery reconciliation — every delivery=required feature in docs/releases/<phase>/features.md (machine-bound via <!-- bubbles:feature id=… spec=… delivery=required --> + a bubbles:reconciled-packet header) MUST map to a TERMINAL + VALIDATE-certified spec; promised-but-unspecced / non-terminal / blocked / implement-self-certified required feature = finding; WARN-grandfathered without the header, BLOCKING with header or --require-coverage (goal/sprint release-phase convergence); malformed reconciled packet fails loud; framework source EXEMPT; compile-time twin enforces rootOutcome.targetReleasePacket coverage in scenario-compile-lint.sh |
release-delivery-reconciliation-guard.sh |
| G110–G126 | Release-train + upkeep + propagation + incident + framework gates: release-train discipline, flag-default-off-on-other-trains, backup / restore-drill / BCDR evidence, env-pollution isolation (G115), offsite-backup-required, audit-trail immutability, backup-retention / PII-classification / secret-rotation declared, propagation policy / validation / ledger, incident severity declared, framework-health evidence, model-tier floor | release-train-guard.sh, env-pollution-scan.sh, propagation-policy-guard.sh, model-tier-advisory.sh + targeted guards |
| G127 | Capability-consumer freshness (framework-source self-validation: every state: shipped capability in bubbles/capability-ledger.yaml declares a non-empty consumers: list whose every path exists on disk; blank-only list = ORPHAN, stray blank entry = MALFORMED, missing path = DANGLING; partial/proposed/deprecated exempt; repos with no ledger no-op; missing yq fails CLOSED only when a ledger is present; the G029 integration-completeness standard applied to the framework's own ledger) |
capability-consumer-freshness.sh |
Canonical Test Taxonomy (binding)
| Category | Means | Real-stack? | Mocks permitted? |
|---|---|---|---|
unit |
Isolated logic, fast | No | Yes (internal) |
functional |
Real-dep functions, may touch DB/filesystem | Optional | External only |
integration |
Multi-component, real deps, real test DB | Yes | External only |
ui-unit |
Component tests with mocked backend | No | Yes (backend) |
e2e-api |
API workflow vs live stack | Yes | External only |
e2e-ui |
UI workflow vs live stack | Yes | External only |
stress |
Burst load vs live stack | Yes | External only |
load |
Sustained load vs live stack | Yes | External only |
A test that uses page.route(), context.route(), cy.intercept(), msw, nock, wiremock, or any internal-service mock and claims to be integration / e2e-api / e2e-ui is misclassified. Reclassify it as ui-unit (or unit) and write a real live-stack test to fill the gap. The framework treats this misclassification as blocking.
Adversarial regression for bug fixes
Bug-fix regression tests MUST include at least one adversarial test case — input that would FAIL if the bug were reintroduced. Tautological regressions (all fixtures already satisfy the broken code path) do not count. Enforced by regression-quality-guard.sh --bugfix.
Resolving a gate failure
- Note the gate ID printed by the guard.
- Open the enforcing script in
bubbles/scripts/and read the failing check's prose. - Fix the underlying gap (in the spec, scope, evidence, test, or code).
- Re-run the guard. Do not change the test/assertion to silence the failure.
Authoritative modules
agents/bubbles_shared/quality-gates.md— full gate catalog with rationaleagents/bubbles_shared/state-gates.md— state-claim integrity gatesagents/bubbles_shared/test-fidelity.md— live-stack authenticityagents/bubbles_shared/e2e-regression.md— persistent regression expectationsbubbles/workflows.yaml— gate-to-mode wiringbubbles/scripts/gate-id-grep.sh— surface every gate ID referenced in the repo