maintain-ci

oleh nvidia

Memelihara dan meninjau alur kerja GitHub Actions NeMo Relay dengan izin per-job yang eksplisit, SHA aksi yang ditetapkan, caching deterministik, alur kerja yang dapat digunakan kembali…

npx skills add https://github.com/nvidia/nemo-relay --skill maintain-ci

Maintain GitHub Actions CI

Companion Guidance

Use karpathy-guidelines alongside this skill for implementation or review work. Keep changes scoped, surface assumptions, and define focused validation before editing.

Use this skill when a change touches .github/workflows/*.yml or .github/workflows/*.yaml, or when reviewing CI behavior for security, reliability, or reproducibility.

Standards

  • Put permissions: on each job that needs token access.
  • Avoid workflow-level permissions unless the repository intentionally centralizes them and the inheritance tradeoff is documented.
  • Keep third-party actions pinned to full commit SHAs and preserve the readable version comment after the SHA.
  • Prefer action-native or ecosystem-native caching over generic actions/cache.
  • Use lockfiles or dependency manifests to drive cache invalidation.
  • Keep deploy and publish permissions isolated to the jobs that need them.
  • Read both caller and callee when a workflow uses workflow_call.
  • Put release-tag validation in the earliest practical caller job when the pipeline has tag-based publish behavior.
  • Keep release-tag policy aligned with RELEASING.md: raw SemVer tags only, no leading v.
  • Keep Codecov component paths aligned with new crates, packages, and generated outputs. Dynamic plugin SDK/protocol paths belong in the plugin component.
  • Keep pure-Python plugin SDK packaging as a single wheel artifact instead of duplicating it across every platform matrix entry.

Permission Model

  • contents: read is the default minimum for checkout-based build, test, docs, and packaging jobs.
  • pull-requests: read is required for PR metadata lookup jobs.
  • pages: write and id-token: write should be limited to Pages deployment jobs and any caller that invokes them through a reusable workflow.
  • For reusable workflows, the caller must grant every permission the called jobs require. The callee cannot elevate beyond what the caller provides.

Caching

  • Prefer astral-sh/setup-uv cache support with cache-dependency-glob anchored to uv.lock.
  • Prefer Swatinem/rust-cache with explicit shared-key and workspaces instead of ad hoc target-directory caching.
  • Avoid caching generated outputs that can hide stale behavior unless the repo already relies on them deliberately.

Review Checklist

  • Each job has the minimum permissions it needs
  • Reusable workflow callers grant only the scopes their callees require
  • Every external action is pinned to a full SHA
  • Cache settings are tied to lockfiles, manifests, or explicit tool versions
  • Secrets are only passed to the jobs that consume them
  • Codecov upload counts match codecov.yml after adding or removing upload jobs
  • Package artifacts include any first-class SDK packages introduced by the change
  • Concurrency, branch filters, and publish guards still reflect release intent
  • Artifact upload, download, and Pages deploy steps have matching permissions
  • Tag-triggered release workflows fail early when a tag violates repo policy

Validation

Start with the narrowest useful checks:

ruby -e 'require "yaml"; Dir[".github/workflows/*.{yml,yaml}"].each { |f| YAML.load_file(f) }; puts "yaml-ok"'
uv run pre-commit run --files .github/workflows/ci.yaml .github/workflows/ci_python.yml

Use ripgrep to inspect the workflow graph before editing:

rg -n "uses:|permissions:|workflow_call|secrets:|upload-artifact|download-artifact|upload-pages-artifact|deploy-pages|codecov|cache" .github/workflows

If local lint passes but the question is whether GitHub will authorize the run, inspect GitHub's permission model and the upstream action or reusable workflow source instead of assuming local success proves remote success.

Canonical References

  • .github/workflows/ci.yaml
  • .github/workflows/ci_python.yml
  • RELEASING.md
  • .pre-commit-config.yaml
  • maintain-packaging
  • validate-change
  • maintain-dynamic-plugins

Lebih banyak skill dari nvidia

compileiq-debug
nvidia
Gunakan ketika ada yang salah: Search() menggantung, semua evaluasi mengembalikan INVALID_SCORE, skor tidak kunjung membaik, setiap konfigurasi mengembalikan angka yang sama, error ptxas…
create-github-pr
nvidia
Buat pull request GitHub menggunakan gh CLI. Gunakan saat pengguna ingin membuat PR baru, mengirimkan kode untuk ditinjau, atau membuka pull request. Kata kunci pemicu -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Memindai isu terbuka lainnya untuk menemukan isu yang mungkin juga diperbaiki atau secara tidak sengaja dirusak oleh suatu PR tertentu. Menghasilkan peluang perbaikan yang berdekatan dan risiko kontradiksi dengan file:baris…
fhir-basics
nvidia
Mengajarkan agen cara kerja API FHIR R4, sumber daya apa saja yang tersedia, cara melakukan kueri dengan parameter pencarian, dan cara mengurai semua format respons dengan benar…
compileiq-validate-result
nvidia
Gunakan SETELAH Pencarian selesai dan SEBELUM mengklaim percepatan atau mengirim ACF. Muat CSV dump_results, ekstrak kandidat top-K (tujuan tunggal)…
changelog-audit
nvidia
Audit Warp CHANGELOG.md sebelum rilis: pulihkan entri yang hilang, urutkan berdasarkan dampak pengguna, perbaiki bahasa entri, bungkus baris, dan (mode cabang rilis) naikkan bandingkan…
maintain-dynamic-plugins
nvidia
Mempertahankan pemuat plugin dinamis NeMo Relay, manifes, SDK asli Rust, protokol pekerja gRPC, SDK pekerja Python, dokumen, pengujian, dan cakupan alur kerja rilis
dgx-diagnose
nvidia
Diagnosis masalah umum DGX Station GB300 — crash CUDA, penargetan GPU yang salah, bug kontainer vLLM/SGLang, masalah status MIG, kesalahan NVLink/Fabric Manager,…