Artifacts carry evidence, status, and non-conclusions together.

Primary artifacts

The main review flow starts with Sig0 and ends with a PR comment summary. Each step keeps claim level and measurement status visible.

• Sig0 output archsig-sig0-v0: component list, dependency edges, signature axes, and metric status for one revision.
• Validation report component-universe-validation-report-v0: duplicate, closure, external target, and policy status checks.
• Snapshot and diff signature-snapshot-store-v0 and signature-diff-report-v0: revision persistence and before / after changes.
• AIR and review reports aat-air-v0, theorem precondition check, Feature Extension Report, policy decision, and PR comment summary.

Surface artifacts

The artifact catalog should be read by surface. Core artifacts describe a bounded repository observation; Review artifacts project that observation into PR review; SFT artifacts project a proposed change into a bounded forecast report; Operational artifacts store calibration and feedback evidence.

• Core artifacts Sig0, validation reports, snapshots, diff reports, metric status, and unmeasured axis records.
• Review artifacts AIR, AIR validation, theorem precondition check, Feature Extension Report, policy decision, PR comment summary, and baseline suppression.
• SFT artifacts ArtifactDescriptor, OperationSupportEstimate, ForecastConeSkeleton, ConsequenceEnvelope, ForecastCalibrationHook, and their validation reports.
• Operational artifacts PR history datasets, feature datasets, outcome linkage, daily ledger, calibration, threshold, ownership, repair adoption, incident correlation, and hypothesis refresh records.

Policy and registry

Policy, registry, and schema artifacts let teams add local thresholds and adapters without claiming that a policy pass is architecture lawfulness.

• Organization policy Warn / fail / advisory rules plus formal claim promotion boundaries.
• Law policy templates and custom rules Reusable policy shape with theorem refs, boundary refs, and non-conclusions.
• Schema compatibility report Field mappings, deprecated fields, required assumptions, and preserved non-conclusions.

Datasets and feedback

Dataset artifacts are empirical records. They connect PR metadata, reports, and outcome observations without turning correlation into causality.

• Empirical datasets PR metadata, before / after signatures, feature reports, theorem checks, and outcome linkage.
• Operational feedback Daily ledger, calibration review, team threshold, ownership boundary, repair adoption, incident correlation, and hypothesis refresh artifacts.

SFT forecasting

Forecasting artifacts bound the input, operation support, finite support, horizon, unknown remainder, and review recommendations for a proposed change.

• ArtifactDescriptor Source refs, action class candidates, scope, missing evidence, measurement boundary, and forecast non-conclusions.
• OperationSupportEstimate Candidate operation families, policy constraints, known forbidden support, unknown remainder, and confidence boundary.
• ForecastConeSkeleton and ConsequenceEnvelope Bounded path candidates, affected regions, comparable axes, obstruction candidates, missing boundary, and review / CI recommendations.

Diagnostic reading

Single-revision diagnostics cover static cycles, SCC size, max depth, fanout, reachable cone, boundary violations, abstraction violations, runtime exposure, and relation complexity. Diff diagnostics cover worsened axes, improved axes, unmeasured axes, evidence diff, and PR attribution candidates.