SPDX-License-Identifier: AGPL-3.0-or-later

Commercial license available

© Concepts 1996–2026 Miroslav Šotek. All rights reserved.

© Code 2020–2026 Miroslav Šotek. All rights reserved.

ORCID: 0009-0009-3560-0851

Contact: www.anulum.li | protoscience@anulum.li

SCPN Quantum Control — Actions History Dashboard

Actions History Dashboard

This page documents the read-only workflow-history audit layer used to keep GitHub Actions history interpretable without deleting evidence for unresolved defects.

The audit helpers classify workflow runs only. They never cancel runs, delete runs, close issues, edit branches, or mutate GitHub state.

Dashboard commands

Run the full Actions history classifier:

python tools/audit_github_actions_history.py \
  --repo anulum/scpn-quantum-control \
  --limit 500

Write the same report as JSON:

python tools/audit_github_actions_history.py \
  --repo anulum/scpn-quantum-control \
  --limit 500 \
  --json > actions-history-audit.json

Run the Link Check history classifier:

python tools/audit_link_check_history.py \
  --repo anulum/scpn-quantum-control \
  --limit 500

Use a captured gh run list fixture instead of querying GitHub:

gh run list \
  --repo anulum/scpn-quantum-control \
  --limit 500 \
  --json databaseId,conclusion,status,workflowName,headSha,headBranch,createdAt,event \
  > actions-history.json

python tools/audit_github_actions_history.py --input actions-history.json
python tools/audit_link_check_history.py --input actions-history.json

Scheduled artefacts

The Actions History Audit workflow runs weekly and can also be triggered manually. It uploads:

Artefact	Meaning
actions-history.json	Raw gh run list output for the sampled history window.
actions-history-audit.json	Classified run history with buckets, reasons, and safe-delete candidates.

The workflow is intentionally read-only. It uses actions: read and contents: read permissions and uploads audit artefacts even when unresolved history is detected.

Classification buckets

Bucket	Meaning	Safe to delete?
clean_success	Completed successfully and retained as current evidence.	No
in_progress	Still running or not completed.	No
resolved_failure	Failed run with a later successful run for the same workflow and branch.	Candidate
unresolved_failure	Failed run without later successful evidence.	No
superseded_cancelled	Cancelled run with later successful evidence for the same workflow and branch.	Candidate
unresolved_cancelled	Cancelled run without later successful evidence.	No
other_completed	Completed with a non-standard conclusion.	No

The Link Check helper uses the same principle, with link-specific buckets:

Bucket	Meaning	Safe to delete?
clean_success	Successful Link Check evidence.	No
resolved_link_failure	Link Check failure followed by a later successful Link Check on the same branch.	Candidate
live_link_failure	Link Check failure with no later successful evidence.	No
accepted_external_transient	Explicitly accepted external-service failure such as rate-limit or HEAD-request blocking.	No
superseded_cancelled	Cancelled Link Check followed by later successful evidence.	Candidate
unresolved_cancelled	Cancelled Link Check without later successful evidence.	No

Safe-delete rule

A failed or cancelled workflow run is only a deletion candidate when all of these are true:

1. The run is completed.
2. The run is classified as resolved_failure, resolved_link_failure, or superseded_cancelled.
3. A later successful run exists for the same workflow and branch.
4. The run is not the only available evidence for an unresolved defect.
5. The deletion is performed manually after reviewing the audit report.

The helpers deliberately do not execute gh run delete.

Accepted external-transient Link Check failures

Accepted transient failures should be explicit and narrow. Use JSON like:

[
  {
    "databaseId": 123456789,
    "reason": "External journal returned HTTP 503 to anonymous HEAD requests."
  }
]

Then run:

python tools/audit_link_check_history.py \
  --input actions-history.json \
  --accepted accepted-link-failures.json

This keeps a visible distinction between real live link failures and external services that intermittently reject automated checks.

Release-safety boundary

The dashboard is a release-safety aid, not a replacement for engineering judgement. A clean dashboard means the sampled history window has no unresolved buckets under the classifier rules. It does not prove that every possible historical defect has been fixed unless the sampled window covers the entire relevant history and the latest CI/security/link-check runs are green.