본문으로 건너뛰기
cccskills
무료GitHub에 공개

ci-spend

The CI cost rules ($100/week of real money, all CI). Use when editing .github/workflows, adding a schedule or cron, dispatching a workflow, publishing a package version, hitting a billing block, or seeing a CI SPEND OFF TRACK line, ci-breaker issue or spend alert.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

연동 도구GitHub
설치 방법 보기

포함된 파일(1)

  • SKILL.md4.7 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

<!-- SYNCED COPY — do not edit here. Canonical: common-docs/skills/ci-spend/SKILL.md This file is distributed to every consuming repo by common-docs/meta/scripts/sync_skills.py. Edit the canonical, run the sync, and commit each repo. Edits made here are overwritten and lost. -->

ci-spend

<!-- rules:stamp -->

Rules from common-docs/policies/ci-spend.md — edit there, never here (python3 common-docs/meta/scripts/one_home.py --write).

  1. Ceiling: $100 a week of real money for building and shipping software. One total, never one service judged alone:
    • GitHub Actions actually billed (netAmount; organization AI-Matrix-Engine and personal armanisadeghi), never list price;
    • RunsOn's AWS cost (Cost Explorer, tag stack=runs-on);
    • CI subscriptions as a weekly share, 52 weeks a year (monthly x 12 / 52, yearly / 52): GitHub Enterprise seats (count and price read from GitHub), GitHub Pro, the RunsOn license (from its start date). Each line shows its $/week and its share of the total. Every price and the cap are billing.ci_spend.* knobs; no lane raises the cap. List price only forecasts when the free GitHub minutes run out: that is a future-billing alarm, never "off track".
  2. Cadence. Only aidream and ai-matrx ride the ~30-minute train. Every other app releases on meaningful change. Matrx Desktop: macOS builds only on Arman's Mac (never on GitHub), Windows/Linux at most twice a week. Matrx Local: paused, no builds. Overnight releases run hourly when main changed. No package version without a real content change. Waste to hunt first: cascading package republishes from small changes.
  3. Every workflow change states its cost. Each workflow has an entry in the repo's .github/ci-cost.json (trigger, runs/week, minutes, OS, $/week, benefit). check_workflow_runner_cost.py warns on a missing or stale entry and fails only an undeclared paid runner (macOS, Windows or larger without a manifest leg or allowlist reason). It never judges spend and has no per-repo share: the weekly total of real money against $100 is judged only by the server spend watch.
  4. Cost without benefit is a defect. Recognize it by these shapes:
    • Red runs repeated without a fix. The ci-breaker label pauses a workflow after 3 reds in a row.
    • Mac or Windows runners on push.
    • A schedule more frequent than the thing it watches changes.
    • Republishing or rebuilding unchanged artifacts.
    • Duplicate workflows doing one job (in one repo or across repos).
    • Polling or waiting on a paid runner.
    • A job with no timeout-minutes.
    • Manual dispatch loops. Every dispatch carries requested_by.
  5. Never work around a billing or spending block. No moving the job to another repo, account or runner. Stop and tell Arman in plain English.
  6. A spend report is the weekly total, then each component with its $/week and % of total, then the waste found and fixed (cascading package republishes from small changes first). The agent-facing notice stays silent unless the 7-day GitHub list-price figure (week_usd) is at least $200; below that only Arman and the owner session see alarms.
<!-- rules:stamp-end -->

The law is ci-spend.md; read it, it is short. How to apply it:

  1. Before touching a workflow, write its .github/ci-cost.json entry first (trigger, runs/week, minutes, OS, $/week, benefit). Run python3 scripts/check_workflow_runner_cost.py. If $/week has no benefit you can name, do not add the workflow.
  2. Prefer paths: filters, Linux runners, timeout-minutes, concurrency with cancel-in-progress, and a path or content-hash check before publishing.
  3. Dispatching a workflow: pass requested_by; never loop a dispatch; if a run goes red, fix it before dispatching again.
  4. Billing or spending block: stop. Do not move the job elsewhere. Tell Arman in plain English.
  5. Seen cost without benefit (any shape in the policy): fix it if it is yours, else one plain sentence to Arman: what, roughly how much, no benefit.
  6. Check spend: GET /ci-spend/status (the weekly total, each component's $/week and share, the cap), or the one-liner in the policy for GitHub alone. Off track means the weekly total of real money is over the cap; free minutes running out is future billing.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Admin debug system wiring for the floating AdminIndicator's live debug data and Copy Full Context. Use when adding debug visibility to a route or feature, wiring useDebugContext or the debug panel, capturing console errors, or enabling the copy-context workflow for a page.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

armanisadeghi/ai-matrx42026년 10월 23일 업데이트

Compact two-icon Copy / Copy-for-AI controls (components/agent-copy). Use when adding copy buttons to a row, card, list, or record; merging duplicate Copy/JSON/AI controls; continuing the copy rollout; or writing a Copy-for-AI payload. NOT for markdown content actions (use rich-document-actions).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

armanisadeghi/ai-matrx42026년 10월 23일 업데이트

Disclosing a surface's existing fixed AI jobs in the shell's top Agents menu. Use when a page, panel, overlay, or window already runs a mandate behind a button, assist, automatic action, or mode; when the agent-disclosure guard names a file; or during a surface check.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

armanisadeghi/ai-matrx42026년 10월 23일 업데이트

The method for getting a cheaper model to match an agent's big-model answers by fixing instructions, inputs and setup. Use when asked to move an agent to a smaller or cheaper model, prove one model beats another, or compare models on an agent's test cases.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

armanisadeghi/ai-matrx42026년 10월 23일 업데이트

The watch-fix-rerun method for making an existing platform agent or automated process efficient and correct: baseline its ledger, run one unit yourself, fix the class behind every wasted call, rerun, record. Use when asked to improve, optimize, watch, or 'make efficient' an agent, a sandbox session, a sync job, or any recurring automated process.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

armanisadeghi/ai-matrx42026년 10월 23일 업데이트

Redux state for agent execution and firing agent shortcuts. Use when editing features/agents/redux/, building agent UI, creating a conversation, touching assembleRequest or NDJSON stream state, wiring agent-state selectors, adding a per-conversation capability, or triggering a shortcut from a button, menu, or code (useShortcutTrigger, launchAgentExecution).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

armanisadeghi/ai-matrx42026년 10월 23일 업데이트

armanisadeghi의 스킬 모두 보기

이 스킬의 문제 신고