Checklists
requirements.md
Specification Quality Checklist: Assertive Test Suite Sanitation
Reviewed: 2026-08-10 Result: PASS
Content Quality
- ✅ No implementation framework or code-structure design leaks into stakeholder requirements.
- ✅ Focuses on maintainer outcomes and trustworthy defect detection.
- ✅ Written for maintainers and reviewers.
- ✅ Every mandatory section is complete.
Requirement Completeness
- ✅ No unresolved clarification markers remain.
- ✅ Requirements are testable and unambiguous.
- ✅ Success criteria are measurable.
- ✅ Functional, non-functional, and constraint requirements are separated.
- ✅ Every requirement has a stable ID and populated status.
- ✅ Non-functional requirements include measurable thresholds.
- ✅ Acceptance scenarios cover primary, failure, duplicate, structural, and CI-routing flows.
- ✅ Edge cases include slow unique guards, platform coverage, migration age, setup failures, and collection attribution.
- ✅ Scope and non-goals are explicit.
- ✅ Assumptions, dependencies, and tracker references are identified.
Mission Readiness
- ✅ Every functional requirement has observable acceptance evidence.
- ✅ User scenarios are independently testable.
- ✅ Outcomes distinguish causal coverage from line-count and marker proxies.
- ✅ The deletion rubric prevents both green-washing and blanket preservation.
- ✅ Branch target and PR-only delivery are explicit.
Validation Notes
- Pre-spec squad:
reviewer-renata,randy-reducer, anddebugger-debbieconverged on evidence-led assertive deletion. The original agent transcripts were not persisted; the later evidence audit and this limitation are recorded ingrounding-report-2026-08-10.md. - Post-spec squad blockers were resolved: contract/architectural gates are unconditional; evidence granularity is family-defined with node expansion on divergence; flake/timing matrices are finite; migration retirement requires compatibility authority; and
FIX_*states cannot close the mission. Post-plan authority review corrected P0 handling to the accepted red-main ADR: exactly one live blocking red per accepted P0, separately accounted from release authority. - Live baseline: 37,444 nodes collected in 94.26 seconds; repeated full-suite start exposed #3283 before test bodies executed.
- Post-plan squad: all reducer, reviewer, and architect findings resolved. The plan uses a lightweight global census, class-specific deep proof, mission-local audit tooling, WP-owned evidence shards, an explicit #3283 lease/atomic-publish protocol, and typed environment/route/workload evidence.
- Healthy base run after manual prewarm: 37,298 passed, 24 failed, 2 errored in 1,689.48 seconds. #2782 was existing; #3284 records the other 23 failures and two errors before baseline acceptance.
- Bulk-edit classification: not applicable; the mission adjudicates heterogeneous tests rather than replacing one repeated identifier/string.