RUNNER-NEUTRAL · METADATA-FIRST
Remove unnecessary CI runs before you switch runners.
A fixed-scope GitHub Actions cost and reliability audit for teams with duplicate triggers, reruns, cancellations, and long queues. Start with exported run metadata—no GitHub App and no source access.
US$250 refundable reservation · US$750 total founding price · four slots
A bounded sample, not a made-up case study.
We exercised the analyzer against the latest 100 public workflow-run records from elizaOS/eliza on 1 August 2026. GitHub reported a capped 40,000-run total, so the figures above describe only that sample.
Metadata records inspected locally
The export contained run IDs, workflow names, commit SHAs, event types, attempts, timestamps, branches, and conclusions. It did not contain repository source or secrets.
Dollar savings claimed
Workflow-run metadata can identify patterns. It cannot prove billed savings. A paid audit reconciles candidates against customer-authorized job and billing exports before assigning financial impact.
How “repeated” is defined
Two or more runs sharing workflow ID, commit SHA, event, and branch. Scheduled or maintenance workflows can repeat intentionally; every group is reviewed before it is classified.
What the founding audit includes.
- 01
Baseline
Local analysis of exported workflow-run metadata, with sample boundaries and data-quality gaps disclosed.
- 02
Human classification
Review retry, duplicate-trigger, cancellation, queue, schedule, and matrix candidates so intentional work is not called waste.
- 03
Financial reconciliation
Map supported findings to redacted job and billing data you choose to share. No impact figure is published without that evidence.
- 04
Change plan
Prioritized workflow changes, acceptance checks, rollback notes, and expected measurement. Customer-authorized implementation can be scoped separately.
- 05
30-day scorecard
A before/after measurement sheet for run volume, failures, cancellations, queue time, and billed usage.
This is not another runner migration.
WarpBuild, RunsOn, Blacksmith, Depot, and others can lower compute prices or accelerate jobs. A CI Waste Audit answers an earlier question: should this work be running at all?
| CI Waste Audit | Runner vendor | |
|---|---|---|
| Primary lever | Remove unnecessary work | Run work cheaper or faster |
| Migration required | No | Usually |
| Source access to start | No | Varies |
| Vendor recommendation | Runner-neutral | Own platform |
Bring an export, not the keys.
The starting audit runs locally against metadata you export. We never ask for passwords, personal access tokens, billing credentials, or a GitHub App installation.
- Source code not required for triage
- Secrets and environment values excluded
- Raw exports deleted within 30 days after delivery
- Customer approves any deeper access separately
FOUNDING ENGAGEMENT · FOUR SLOTS
Start with a refundable reservation.
US$250 reserves the audit. If the metadata triage does not find at least one actionable waste class worth deeper review, the reservation is refunded. If it proceeds, US$500 is due for the completed report and 30-day scorecard.
No guaranteed savings. Scope is confirmed before work begins. Read the terms and privacy notice.