stark-gha-cost

Diagnose and reduce GitHub Actions and GHAS billing across organizations and enterprises.

Updated Mar 16, 2026
One-click install
npx skills add https://github.com/21StarkCom/stark-skills --skill stark-gha-cost-21starkcom
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: stark-gha-cost
Source: https://github.com/21StarkCom/stark-skills/tree/main/runtime-overrides/codex/skill/stark-gha-cost
Command: npx skills add https://github.com/21StarkCom/stark-skills --skill stark-gha-cost-21starkcom

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? GitHub bills two unrelated meters — Actions minutes and GHAS seats — and teams routinely optimize the wrong one, chase billing anomalies with workflow edits, or pay for runaway matrices and poor caches. This Skill traces spend from product to repository, workflow, and run so fixes target the actual cost driver. ## Core Features & Use Cases - Spend breakdown: Runs scripts that rank billing by product, SKU, and repository using the enhanced-billing usage API, plus GHAS committer-seat consumption for Secret Protection and Code Security. - Per-repo drill-down: Classifies a repo's driver as high run volume, matrix fan-out, long runs, or a billing anomaly, with per-job rounding and OS multiplier analysis (Linux 1x, Windows 2x, macOS 10x). - Ranked fixes shipped via PR: Applies levers like concurrency cancel-in-progress, path filters, cron reduction, cache warming, and seat right-sizing through branch-and-PR workflows, and drafts GitHub Support tickets for genuine mis-billing. - Use Case: An organization sees a surprise GitHub bill; the Skill identifies that one private repo's test matrix accounts for 90% of Actions minutes, then ships a PR that shards jobs and shares build caches. ## Quick Start Ask the assistant to diagnose why your GitHub Actions bill is high for your enterprise or organization and propose cost-reduction fixes.

Frequently Asked Questions about stark-gha-cost

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find out why my GitHub Actions bill is so high?▼

Run the cost breakdown script with an admin:enterprise or admin:org PAT to rank spend by product, SKU, and repository via the enhanced-billing usage API. Then drill into the top repo to classify the driver as run volume, matrix fan-out, long runs, or a billing anomaly.

How do I reduce GitHub Actions minutes usage?▼

Apply ranked levers: add concurrency cancel-in-progress on PR workflows, path-filter expensive macOS jobs, cut over-frequent cron schedules, and shrink test matrices by bucketing jobs and sharing build caches. Every job rounds up to a full billed minute, so many short jobs multiply cost.

Do public repositories pay for GitHub Actions minutes?▼

No. Public repositories get unlimited free GitHub-hosted Actions minutes and storage, so a busy public repo costs nothing regardless of run count. If a public repo's bill looks high, the cost is a downstream resource like registry storage or deployed cloud services.

What is the difference between Actions minutes and GHAS seats billing?▼

GitHub bills two separate meters: Actions minutes for compute and GHAS per-committer seats (Secret Protection around $19, Code Security around $30 monthly). Self-hosting runners only affects Actions minutes and never reduces the seat bill; Dependabot and dependency graph are free.

Why does my GitHub bill show huge minutes when my runs are short?▼

When the timing API reports near-zero billable time but the bill shows large minutes, that is a billing anomaly that workflow edits cannot fix. The Skill drafts a GitHub Support ticket for mis-billing while still applying legitimate optimizations to any real usage.

When do self-hosted runners actually save money on GitHub Actions?▼

Self-hosted runners are marginal if you must provision new compute, but are the biggest lever when you already own a cluster, such as ARC ephemeral runners on existing GKE. They never apply to macOS jobs or public repositories, and they do not reduce GHAS seat costs.