AUTO-MAS-Project/AUTO-MAS / .agents/skills/mas-ci-checks
mas-ci-checks
VerifiedUse when an AUTO-MAS PR is red in CI, when reproducing a dev-PR gate locally before pushing, when a workflow needs a skip-changelog or changelog-maintenance label, or when editing .github/workflows, .github/actions, scripts/changelog.py or scripts/check_reverts.py. Covers every gate that runs on a PR to dev and the exact local command for each.
MAS CI Checks
Objective
Turn the dev-PR gates into local commands, and decide "our change is wrong" versus "runner or mirror infra is wrong" before touching code.
Gate Map (PR with base dev)
format-check and check-reverts are filtered to base dev; check-changelog, check-uv-lock and ci-package run on every PR, including release/* cherry-pick PRs, where check-changelog also adds --dev-ref origin/dev to reject dev-only commits.
| Workflow | Job name | What actually runs | Local reproduction |
|---|---|---|---|
format-check.yml | 格式检查 | uv sync --locked then uv run --no-sync ruff format --check .; in frontend/: yarn lint (oxlint), yarn typecheck (vue-tsc), yarn format:check (oxfmt) | see Local Reproduction |
check-changelog.yml | 检查更新日志碎片与版本号 | fetches base, dev and all tags, then python scripts/changelog.py check --pr-base origin/<base> --pr-kind normal plus --skip-changelog / --maintenance when the PR carries the matching label | git fetch origin dev --tags, then the same command with --pr-base origin/dev and the same label flags; omitting a flag the PR has gives a false red |
check-uv-lock.yml | 检查 uv.lock 与依赖清单一致 | uv lock --check | same |
check-reverts.yml | 检查是否删掉了他人近期合入的代码 | blames every deleted line; lines another author merged within 14 days are annotated in Files changed as a hint only, it never fails. Skipped entirely for Release v* PRs | not needed; the job is informational |
label-changelog.yml | 同步碎片标签 | python scripts/label_changelog.py, continue-on-error: true | not needed; failure does not judge the PR |
ci-package.yml | 构建未签名 Windows 验证包 | real windows-latest build of frontend + Electron + backend, ~4 minutes; skipped for **.md / docs/**-only diffs | expensive; only re-run locally when the change touches Electron or packaging |
sync-cnb.yml | sync | mirrors the repo to cnb.cool after push / branch create | gh run rerun <run-id>; never a code problem |
build-app.yml, prepare-release.yml, absorb-changelog.yml, mirrorchyan*.yml, weekly-format.yml run on main / release/* / release PRs / schedules. They are not PR gates; do not judge a feature PR by them.
Local Reproduction
- Python format:
uv sync --locked, thenuv run --no-sync ruff format --check <files>; if it fails,uv run --no-sync ruff format <files>. Always go through uv: the ruff version is pinned byuv.lock(uv run ruff --version). A system-wide ruff formats differently, so the local run passes while CI stays red. - Frontend:
cd frontend && yarn lint && yarn typecheck && yarn format:check. yarn typecheckonly runsvue-tsc -p tsconfig.app.json --noEmit, so it does not coverfrontend/electron/. For Electron main-process changes (electron/services,electron/ipc) also runnode node_modules/typescript/bin/tsc -p tsconfig.electron.json --noEmitfromfrontend/.- Changelog fragment: for a user-visible change run
python scripts/changelog.py add <type> <project-key> "<sentence>"; for genuinely user-invisible work (docs, CI, tests, invisible refactor) label the PRskip-changelogwithgh pr edit <n> --add-label skip-changelog. The workflow also triggers onlabeled/unlabeled, so no extra empty commit is needed.changelog-maintenancemaps to--maintenanceand only lets a maintainer reorganizeCHANGELOG.mddirectly (e.g. afterpython scripts/changelog.py sync); never put it on a normal feature or fix PR. - Dependency changes:
uv lock --check. - Format-only repairs: reformatting rewraps lines, so after
ruff formatconfirm withgit diffthat only your own hunk moved.
Rules
- Read the log before editing:
gh run view <run-id> --log-failedgives the exact file and line plus the expected layout. Do not guess the rewrap. - Do not force push to repair CI (root
AGENTS.mdforbids force push outright). Append astyle(...)commit and let the PR be squashed on merge. - Every
uses:/image:reference in workflows, composite actions and CNB pipelines stays pinned to a full 40-character commit SHA or@sha256:digest; seemas-skillsProject Rule Areas item 7. - A
syncfailure such asremote: error: cannot lock ref 'refs/heads/<branch>': reference already existsis mirror infra: rerun the workflow and keep the code untouched. - Never turn CI green by deleting tests, weakening a check, or adding
continue-on-error.
Anti-Patterns
- Reproducing a format failure with the system ruff or oxfmt instead of the
uv.lockversion. - Treating the CNB mirror job as a code failure.
- Labeling a user-visible change
skip-changelogto get past the fragment gate. - Pushing an empty commit after adding a label.
- Assuming
yarn typecheckproves Electron main-process types are sound.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 198k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |