菜单
Run or plan Yuzu's `/test` pre-commit and pre-push validation pipeline with quick, default, full, instructions, and quarantine modes. Use when the user says `/test`, `/test --quick`, `/test --full`, asks for the full test gate, or needs the Yuzu test-runs DB workflow.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Run the Yuzu full governance pipeline on a commit range. Produces Gate 1 Change Summary, then orchestrates Gates 2-6 as parallel agent fan-outs with pre-written prompt preambles. Use when the user says "/governance <range>" or "full governance" or asks to run the multi-gate review pipeline on a PR/commit set.
Authentication & Authorisation control plane for Yuzu — the canonical entry point for any work on RBAC, OIDC SSO, SAML, SCIM, MFA/TOTP, AD/Entra integration, API tokens, session lifecycle, enrollment, and the audit/evidence chain. Use when the user says "/auth-and-authz", "/auth", "/iam", asks to plan or implement an enterprise A&A feature, asks "what's our auth gap to enterprise readiness", asks to audit current auth state against SOC 2 CC6.x / Workstream B, or starts work that touches `auth_*`, `rbac_*`, `oidc_*`, `api_token_*`, `enrollment_*`, or `cert_store.*`. The skill bundles current-state inventory, required-features inventory, gap matrix, the canonical workflow for adding a new A&A feature, and the load order for the routed reference docs.
Cut a Yuzu release end-to-end. Runs preflight checks, validates compose-file version pins, pushes the tag, monitors the release workflow until artifacts publish, troubleshoots known failure modes (artifact download bugs, version mismatch, Windows signing, macOS notarization), verifies the GitHub Releases page has every expected asset including the Compose Wizard zip, and produces a release record. Use when the user says "/release vX.Y.Z" or asks to cut a release.
Run an adversarial two-phase code review of a change with TWO independent reviewers — Claude and Codex — who review alone, then cross-examine each other's findings, then Claude synthesizes a single weighted verdict. Use when the user says "/adversarial-review", "adversarial review", "review this with Codex", "get Codex to review", "two-reviewer review", "cross-examine this PR", or wants a second independent model to grade a change before merge.
Authentication & Authorisation control plane for Yuzu — the canonical entry point for any work on RBAC, OIDC SSO, SAML, SCIM, MFA/TOTP, AD/Entra integration, API tokens, session lifecycle, enrollment, and the audit/evidence chain. Use when the user says "/auth-and-authz", "/auth", "/iam", asks to plan or implement an enterprise A&A feature, asks "what's our auth gap to enterprise readiness", asks to audit current auth state against SOC 2 CC6.x / Workstream B, or starts work that touches `auth_*`, `rbac_*`, `oidc_*`, `api_token_*`, `enrollment_*`, or `cert_store.*`. The skill bundles current-state inventory, required-features inventory, gap matrix, the canonical workflow for adding a new A&A feature, and the load order for the routed reference docs.
Canonical patterns for caching in Yuzu CI workflows. Two snippets — one for ephemeral GHA-hosted runners (split actions/cache/restore + actions/cache/save, never `save-always: true`) and one for self-hosted runners (local filesystem cache under `runner.tool_cache`, no GHA cache round-trip). Use when adding a new vcpkg/ccache/dependency cache step to any workflow under `.github/workflows/`, or when reviewing a PR that touches `actions/cache@`.
| name | test |
| description | Run or plan Yuzu's `/test` pre-commit and pre-push validation pipeline with quick, default, full, instructions, and quarantine modes. Use when the user says `/test`, `/test --quick`, `/test --full`, asks for the full test gate, or needs the Yuzu test-runs DB workflow. |
Use this skill for the high-level Yuzu validation pipeline. Route low-level build choices through $yuzu-build, Windows through $yuzu-windows-msvc, and Erlang caveats through the gateway guidance below.
--quick: build plus offline C++ unit, EUnit, and Dialyzer. No live stack, no synthetic UAT.--full: default plus OTA stubs/flows, sanitizer dispatch, coverage, and perf measurement. About 60-120 minutes.--instructions: live-stack instruction definition suite only.--instructions-quarantine: local quarantine ceremony only. Do not run on remote or SSH-only hosts.--force-cleanup: clean this run's dangling containers before start.--keep-stack: leave debug stacks running when the script supports it.Every full pipeline run records gate results and timings in ~/.local/share/yuzu/test-runs.db, overrideable with YUZU_TEST_DB. Query with:
bash scripts/test/test-db-query.sh --latest
bash scripts/test/test-db-query.sh --last 5
bash scripts/test/test-db-query.sh --diff RUN_A RUN_B
bash scripts/test/test-db-query.sh --trend timing=phase7.perf
bash scripts/test/test-db-query.sh --flaky
Initialize each manual orchestration with a RUN_ID, TEST_DIR=/tmp/yuzu-test-$RUN_ID, LOG_DIR=$HOME/.local/share/yuzu/test-runs/$RUN_ID, COMMIT, BRANCH, and BUILDDIR=$(build_dir) from scripts/test/_portable.sh. Record gates with scripts/test/test-db-write.sh.
Wall-clock order is fixed:
bash scripts/test/preflight.sh or --force-cleanup.rebar3 as prod release, and docker images except in quick mode.bash scripts/test/test-upgrade-stack.sh ...; skipped in quick mode.bash scripts/start-UAT.sh; skipped in quick mode.tests/coverage-baseline.json.Phase 1 is mandatory-blocking. Other phases should normally continue so the operator gets one prioritized report.
Phase 8 must not stop the native UAT stack from scripts/start-UAT.sh. A successful default/full run intentionally leaves a working stack at tested HEAD for human inspection. Only stop it if the user explicitly asks.
scripts/ensure-erlang.sh before direct rebar3 work, then verify command -v erl.rebar3 eunit --dir apps/yuzu_gw/test; the --dir flag is mandatory.rebar3 eunit --module X as isolated; it runs after the full --dir phase in the same VM.REBAR_BASE_DIR values for EUnit and CT to avoid _build races.Perf is measure-only as of 2026-05-03. scripts/test/perf-gate.sh records metrics and exits PASS; humans inspect trend and distribution shape using the test-runs DB and docs/perf-baseline-calibration-2026-05-03.md.
For --quick, resolve the build directory and run only compile plus offline suites. Do not start UAT and do not invoke synthetic UAT.
For --instructions, use scripts/test/instructions-tests.sh after bringing up only the required live stack.
For --instructions-quarantine, use scripts/test/instructions-quarantine.sh only on a local machine where temporary self-disconnect is acceptable.