GitHub Actions SHA Rerun Debt Audit

Audit rerun debt by commit SHA to find commits that repeatedly burn CI minutes across workflows.

Security Scan

The skill does what it says — it parses local GitHub Actions JSON exports and scores rerun debt — but there are small documentation mismatches (it references the gh CLI but doesn't declare it) you should be aware of before running it in your environment.

Purpose & Capability

Name/description match the implementation: the script reads GitHub Actions run JSON files, correlates attempts by run id, aggregates metrics by commit SHA, and emits a ranked report. Required binaries (bash, python3) are reasonable. One inconsistency: SKILL.md shows using the `gh run view` command to collect JSON, but the skill's declared required binaries do not include `gh` (GitHub CLI) nor does it declare any GitHub credentials.

Instruction Scope

Runtime instructions are limited to collecting JSON run exports and running the bundled script against them; the script reads files matched by RUN_GLOB and does local aggregation and reporting. There are no network calls or external endpoints in the shown code. The SKILL.md suggests using `gh run view` (which will contact GitHub and requires authenticated gh), but the script itself only processes local files.

Install Mechanism

No install spec — instruction-only skill with a bundled script. Nothing is downloaded or installed by the skill itself.

Credentials

The skill requests no credentials or environment variables beyond operational parameters (RUN_GLOB, TOP_N, etc.). However, collecting run JSON via `gh run view` (shown in SKILL.md) requires the GitHub CLI and authenticated access to GitHub; those prerequisites are not declared. No other unexpected secrets or config paths are requested by the script.

Persistence & Privilege

The skill is not always-enabled, does not request elevated persistence, and does not modify other skills or system-wide config. It only reads files matched by RUN_GLOB and writes output/exit codes.

Assessment

This skill appears to do exactly what it claims: analyze local GitHub Actions run JSON files and report rerun debt by commit SHA. Before installing/running it: (1) Inspect the full script (you were shown a truncated portion) to confirm no unexpected behavior; (2) note that SKILL.md shows using `gh run view` to collect run JSON — running that requires the GitHub CLI and authenticated access (gh stores credentials or uses your environment), so be mindful of what account/permissions are used when collecting artifacts; (3) the skill itself does not exfiltrate data or contact external endpoints in the shown code, but it will process whatever JSON files you point it at — avoid feeding it sensitive files from unknown sources; (4) run it first against the provided fixtures (RUN_GLOB set to fixtures) to validate behavior in a safe context. If you want higher assurance, request the author to explicitly declare the gh dependency and include full source for review.

Use this skill to detect commits that trigger repeated GitHub Actions reruns and failed outcomes across multiple workflows.

What this skill does

• Reads GitHub Actions run JSON exports
• Correlates attempt history by run id and latest outcome per run
• Aggregates rerun debt by repository + commit SHA
• Scores risk using rerun rate, failed-run count, workflow spread, and wasted rerun minutes
• Emits severity (ok, warn, critical) for CI gates

Inputs

Optional:

• RUN_GLOB (default: artifacts/github-actions/*.json)
• TOP_N (default: 20)
• OUTPUT_FORMAT (text or json, default: text)
• MIN_RUNS (minimum runs per SHA, default: 3)
• WARN_RERUN_RATE (0..1, default: 0.25)
• CRITICAL_RERUN_RATE (0..1, default: 0.45)
• WARN_FAILED_RUNS (default: 2)
• CRITICAL_FAILED_RUNS (default: 4)
• WARN_WASTED_MINUTES (default: 25)
• CRITICAL_WASTED_MINUTES (default: 75)
• WARN_WORKFLOWS (distinct workflows affected, default: 2)
• CRITICAL_WORKFLOWS (default: 4)
• WORKFLOW_MATCH / WORKFLOW_EXCLUDE (regex, optional)
• BRANCH_MATCH / BRANCH_EXCLUDE (regex, optional)
• EVENT_MATCH / EVENT_EXCLUDE (regex, optional)
• REPO_MATCH / REPO_EXCLUDE (regex, optional)
• HEAD_SHA_MATCH / HEAD_SHA_EXCLUDE (regex, optional)
• FAILURE_CONCLUSIONS (comma-separated, default: failure,cancelled,timed_out,startup_failure,action_required)
• FAIL_ON_CRITICAL (0 or 1, default: 0)

Collect run JSON

gh run view <run-id> --attempt <attempt> \
  --json databaseId,runAttempt,workflowName,event,headBranch,headSha,conclusion,createdAt,updatedAt,runStartedAt,url,repository \
  > artifacts/github-actions/run-<run-id>-attempt-<attempt>.json

Run

Text report:

RUN_GLOB='artifacts/github-actions/*.json' \
bash skills/github-actions-sha-rerun-debt-audit/scripts/sha-rerun-debt-audit.sh

JSON output + fail gate:

RUN_GLOB='artifacts/github-actions/*.json' \
OUTPUT_FORMAT=json \
FAIL_ON_CRITICAL=1 \
bash skills/github-actions-sha-rerun-debt-audit/scripts/sha-rerun-debt-audit.sh

Run against bundled fixtures:

RUN_GLOB='skills/github-actions-sha-rerun-debt-audit/fixtures/*.json' \
bash skills/github-actions-sha-rerun-debt-audit/scripts/sha-rerun-debt-audit.sh

Output contract

• Exit 0 in report mode (default)
• Exit 1 when FAIL_ON_CRITICAL=1 and one or more SHA groups are critical
• Text mode prints summary + ranked SHA risk groups
• JSON mode prints summary + ranked groups + critical groups