Skip to content

Artifact reference and generated sample map

Use this page when you need to understand where SDETKit writes evidence, which files are uploaded by CI, and which committed files are examples rather than live run outputs.

SDETKit's artifact story is intentionally evidence-first: gates, investigation, diagnosis, and maintenance lanes produce machine-readable files that humans can inspect before any remediation is considered.

Safety baseline

Investigation, diagnosis, recommendation, planning, proof selection, classification, trending, candidate selection, probation, policy proposal, and dry-run outputs are diagnostic/report-only by default.

A file that says a fix is possible is not approval to mutate the repository. Mutation requires a separate reviewed policy path, explicit guarded flags, PR-only guardrails, and proof artifacts attached to the workflow run or pull request.

Primary local artifacts

Purpose Path Produced by Notes
Release-readiness evidence package build/sdetkit/release-readiness-evidence/package.json, build/sdetkit/release-readiness-evidence/package.md python -m sdetkit.release_readiness_evidence_package --root . --out-json build/sdetkit/release-readiness-evidence/package.json --out-md build/sdetkit/release-readiness-evidence/package.md --format text Reporting-only release evidence bundle; does not authorize publish or merge.
Fast release-confidence gate build/gate-fast.json python -m sdetkit gate fast --format json --stable-json --out build/gate-fast.json First artifact to inspect.
Release preflight gate build/release-preflight.json python -m sdetkit gate release --format json --out build/release-preflight.json Pairs with gate-fast.json for ship/no-ship decisions.
Doctor checks build/doctor.json python -m sdetkit doctor --format json --out build/doctor.json Optional JSON form for machine-readable doctor output.
Adoption surface build/sdetkit/adoption-surface.json python -m sdetkit adoption-surface --root . --out build/sdetkit/adoption-surface.json --format text Read-only repository surface map for languages, package managers, CI, security tools, artifact surfaces, and recommended proof commands. Does not run proof or authorize automation.
Failure investigation build/investigation/failure.json python -m sdetkit investigate failure --log build/quality.log --format json --out build/investigation/failure.json Diagnostic-only triage result.
Repository investigation build/investigation/repo.json python -m sdetkit investigate repo --root . --format json --out build/investigation/repo.json Diagnostic-only surface narrowing.
Adaptive diagnosis build/adaptive-diagnosis.json PYTHONPATH=src python -m sdetkit.adaptive_diagnosis --log build/quality.log --format json --out build/adaptive-diagnosis.json Evidence-fitted diagnosis and proof commands.
Adaptive diagnosis learning .sdetkit/adaptive-diagnosis-memory.jsonl python -m sdetkit adaptive learn record build/adaptive-diagnosis.json --db .sdetkit/adaptive-diagnosis-memory.jsonl JSONL learning events for matched signals, candidates, proof commands, recurrence, and operator outcome feedback.
Adaptive learning summary operator-chosen JSON/stdout python -m sdetkit adaptive learn summarize --db .sdetkit/adaptive-diagnosis-memory.jsonl --format json Rollup of top recurring scenarios, weakest lanes, and promotion/exampletion calibration actions.
Operator onboarding evidence flow build/sdetkit/operator-onboarding-evidence-flow/flow.json, build/sdetkit/operator-onboarding-evidence-flow/flow.md python -m sdetkit.operator_onboarding_evidence_flow --root . --out-json build/sdetkit/operator-onboarding-evidence-flow/flow.json --out-md build/sdetkit/operator-onboarding-evidence-flow/flow.md --format text Reporting-only operator path map; does not authorize remediation, patching, merge, dismissal, or semantic-equivalence claims.
Operator brief build/sdetkit/operator-brief.md python -m sdetkit adaptive brief --gate build/gate-fast.json --diagnosis build/adaptive-diagnosis.json --out build/sdetkit/operator-brief.md One-page handoff with gate result, diagnosis, candidates, first proof command, safe-fix decision, and owner action.
Operator PR comment build/sdetkit/operator-comment.md python -m sdetkit adaptive brief --gate build/gate-fast.json --diagnosis build/adaptive-diagnosis.json --format comment --out build/sdetkit/operator-comment.md Compact PR-safe summary for green, safe mechanical, and review-first unknown flows.

For schema-oriented contracts, see artifact-contract-index.json. That JSON index is the machine-readable contract inventory; this page is the human map.

Release evidence recipes and proof samples

Use Release evidence recipes and sanitized proof samples for copy-ready report commands, collection-state interpretation, freshness checks, and troubleshooting.

Committed contract examples:

These files are sanitized examples governed by focused exact-byte tests. They are not current-run evidence and grant no release, publish, merge, patch, or security-dismissal authority.

Maintenance autopilot upload set

The maintenance-autopilot workflow uploads these files when they are produced:

Category Paths How to read them
Autopilot report build/maintenance/autopilot/autopilot-report.json, build/maintenance/autopilot/autopilot-report.md Start here for the run summary.
Command-center plans build/maintenance/autopilot/command-center-dry-run-plan.json, build/maintenance/autopilot/command-center-live-plan.json Plans are audit evidence; they do not authorize mutation by themselves.
Core signals build/maintenance/autopilot/sdet_check.json, build/maintenance/autopilot/doctor.json, build/maintenance/autopilot/review.json, build/maintenance/autopilot/security-check.json Cross-check release, doctor, review, and security posture.
Investigation and diagnosis build/maintenance/autopilot/adaptive-diagnosis.json, build/maintenance/autopilot/adaptive-diagnosis.md, build/maintenance/autopilot/adaptive-diagnosis-error.json Diagnostic-only failure explanation and proof recommendations.
Safe-fix audit trail build/maintenance/autopilot/safe-fix-plan.json, build/maintenance/autopilot/adaptive-safe-remediation-result.json, build/maintenance/autopilot/adaptive-safe-remediation-result.md, build/maintenance/autopilot/adaptive-safe-remediation-error.json, build/maintenance/autopilot/adaptive-safe-commit-result.json Treat as guarded remediation evidence, not broad auto-fix permission.
Learning memory .sdetkit/maintenance/failure-memory.jsonl, .sdetkit/maintenance/adaptive-safe-fix-memory.jsonl, build/maintenance/autopilot/adaptive-safe-fix-learning-result.json, build/maintenance/autopilot/adaptive-safe-fix-learning-rollup.json, build/maintenance/autopilot/adaptive-safe-fix-learning-error.json Use to understand repeated patterns and remembered outcomes.

The workflow keeps pull-request runs diagnostic by avoiding live remediation flags on pull_request events. Non-PR side effects are separately guarded in the workflow and should be reviewed with the uploaded artifacts.

Committed generated and sample artifacts

docs/artifacts/ is a committed evidence gallery. It contains generated/sample artifacts used for product proof, adoption examples, completion packs, and historical examplenstrations.

Labels to use when reading that tree:

  • Generated/sample material: committed examples under docs/artifacts/ are not produced by your local run unless a command explicitly writes there.
  • Canonical first-proof evidence: current live-adoption proof packs are linked from the README and live-adoption-product-proof.md.
  • Historical completion report packs: dated or numbered completion report folders are preserved for traceability and should not be treated as the current operator runbook.
  • Runtime artifacts: files under build/, .sdetkit/out/, and .sdetkit/maintenance/ are generated by local or CI commands and are the evidence to cite for a specific branch.

Use this table before opening raw logs. It maps the most common artifact signals to the next safe operator action.

If you see... Open this artifact first Next safe action Mutation posture
A local release gate failed build/gate-fast.json or build/release-preflight.json Read ok, failed_steps, and the failing step output; then run sdetkit investigate failure on the captured log. Diagnostic-only.
A CI log or PR check failed build/investigation/failure.json Read classification, summary, next_actions, proof_commands, diagnostic_only, automation_allowed, and requires_human_review. Diagnostic-only.
The owning area is unclear build/investigation/repo.json and the surface Markdown output Use the repo/surface result to choose the smallest focused proof command. Diagnostic-only.
You need to choose proof commands or rollout depth before running gates build/sdetkit/adoption-surface.json Review detected languages, package managers, CI, security tooling, artifact surfaces, and recommended proof commands; then choose the smallest proof path. Read-only, non-authorizing.
A maintenance-autopilot run uploaded artifacts build/maintenance/autopilot/autopilot-report.md and build/maintenance/autopilot/adaptive-diagnosis.md Summarize the diagnosis and proof commands in the PR or issue before considering remediation. Diagnostic-only on pull requests.
A safe-fix plan exists build/maintenance/autopilot/safe-fix-plan.json Review it as an audit plan; verify policy, branch, diff, and proof artifacts before any guarded lane runs. Not approval to mutate.
A pattern keeps recurring .sdetkit/maintenance/failure-memory.jsonl and .sdetkit/maintenance/adaptive-safe-fix-memory.jsonl Use memory to explain recurrence and decide whether a policy proposal is warranted. Evidence for review, not auto-approval.
You are reading committed samples docs/artifacts/ and live-adoption-product-proof.md Treat them as examples or historical proof packs; use runtime build/ or .sdetkit/ files for the current branch decision. Not current-run evidence unless regenerated.

Quick rules:

  1. For ship/no-ship decisions, start with build/gate-fast.json and build/release-preflight.json.
  2. For a failed command or CI log, run sdetkit investigate failure and read diagnostic_only, automation_allowed, safe_to_auto_fix, and requires_human_review.
  3. For recurring maintenance failures, open the maintenance-autopilot upload and read the report, adaptive diagnosis, safe-fix plan, and learning memory together.
  4. For human remediation steps, use remediation-cookbook.md after the investigation or artifact summary points to a specific failure class.
  5. For CI upload interpretation, use ci-artifact-walkthrough.md.

Evidence circuit artifact source map

Use this source map when a PR or workflow run references the completed #1748 through #1761 evidence circuit. It connects the human docs to the artifact surfaces that reviewers should inspect first.

Review question Open first Then inspect Authority posture
What is the completed evidence path? evidence-graph-summary.md evidence-circuit-architecture-checkpoint.md Reporting-only source map
How should a reviewer read the evidence? operator-evidence-review-guide.md PR Quality, Runtime Proof, and ProtectedVerifier artifacts Human review only
Which artifact explains PR-facing evidence? PR Quality action report or review dashboard Dashboard and reporting polish, Runtime Proof summary, and ProtectedVerifier decision Does not authorize merge
Which artifact explains runtime evidence? runtime-proof/summary/runtime-proof-artifacts.json runtime-proof/summary/runtime-proof-artifacts.md Does not authorize patching or dismissal
Which artifact explains authority evidence? PR Quality artifact manifest authority evidence source map in the artifact center Denied authority remains denied
Which artifact explains replay evidence? benchmark replay report or Runtime Proof benchmark section ProtectedVerifier benchmark replay evidence Replay supports review, not approval

Evidence circuit reading order

  1. Start with Evidence graph summary for the reviewer-facing source map.
  2. Use Operator evidence review guide to decide which evidence source needs inspection.
  3. Open PR Quality artifacts for the maintainer-facing summary.
  4. Open Runtime Proof artifacts when runtime evidence or benchmark replay is referenced.
  5. Open ProtectedVerifier output when authority boundaries are disputed.
  6. Return to the PR only after confirming patch application, security dismissal, merge authorization, and semantic-equivalence claims remain denied.

This table is a navigation aid. It does not replace the proof command, PR review, or release-readiness checklist.

After using the evidence circuit artifact source map, continue with:

  1. Dashboard and reporting polish for dashboard reading order.
  2. Evidence circuit review pack for the complete reviewer path.
  3. Release-readiness evidence handoff when evidence must be summarized for release review.

These links are navigation aids only. They do not authorize patch automation, security dismissal, merge, or semantic-equivalence claims.