automation-pipeline-integration
Wire the automation test project into the repository's own pipeline so every task must pass the suite before it can leave QA
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-15
- Stars
- ★ 122
- Forks
- 18
01 前言 Front Matter
- name
- automation-pipeline-integration
- enabled
- false
- category
- testing
- description
- Wire the automation test project into the repository's own pipeline so every task must pass the suite before it can leave QA
03 SKILL.md 正文
Automation Pipeline Integration
Overview
An automation suite that only runs when someone remembers it is decoration. The suite must run in the repository's own pipeline — the one this platform triggers automatically when a task moves to ready_for_qa — so a task cannot pass QA while the suite is red. Each repository runs its own tests in its own pipeline; there is no shared cross-project test job.
How the platform runs it
When a task reaches ready_for_qa, the QA pipeline runs the repo's build and test stages: the repository's test_command override if set, otherwise language auto-detect. A red pipeline sends the task back to need_revision automatically with the failed stage's log; a green pipeline releases the task to QA dispatch. Check results with get_pipeline_status; re-trigger manually via the task's pipelines endpoint after a fix.
Wiring the suite in
- One entrypoint command. Give the automation project a single script that provisions everything and exits non-zero on any failure — e.g.
scripts/e2e.sh: start disposable dependencies (Testcontainers/compose), boot the app from the current branch build, start WireMock stubs, run the suite, tear down. - Make the pipeline call it. Set the repository's
test_commandto run unit tests AND the automation entrypoint (go test ./... && ./qa-automation/scripts/e2e.sh, or the package.json script that does both). If the repo's CI (GitHub Actions etc.) is the source of truth for deploys, add the same entrypoint there too — same command locally, in the QA pipeline, and in CI, so green means the same thing everywhere. - Keep it self-contained. The pipeline environment provides no shared DB and no external network guarantees: the suite brings its own database container (test-database-seeding) and stubs (test-doubles-wiremock). If the environment has no container runtime, the entrypoint must fail loudly with a clear message — never silently skip tests and report green; raise the infra gap on the task instead.
- Keep it fast and deterministic. Budget the suite (minutes, not an hour); parallel-safe, order-independent tests; no sleeps, no shared state. Flaky tests get fixed or quarantined with a task — a suite people rerun until green enforces nothing.
Coverage reporters are mandatory
The platform parses the test job's log for a total coverage percentage and surfaces it on the pipeline and job — but only if the suite prints one in a format it recognizes. A suite with no coverage reporter enabled leaves that number blank, silently, and nobody is told the pipeline stopped proving anything about coverage. Every automation/unit suite wired into test_command must enable one:
- Go:
go test -cover ./...(prints acoverage: NN.N% of statementsline per package), orgo tool cover -func=coverage.outafter a-coverprofile=coverage.outrun (prints thetotal: (statements) NN.N%summary line). - JS/TS (Jest/Vitest): run with
--coverageso the run ends with an istanbul text summary (Lines : NN.N%or theAll files | ... | NN.N |table) in the job's stdout — not only an HTML/lcov file the log never shows. - Python (pytest):
pytest --covwith the default terminal report, which prints aTOTAL ... NN%line.
The report must land in the job's console output (stdout/stderr), because that is the only thing the platform's log-based parser can read — a coverage artifact written only to disk (coverage.out, lcov.info, htmlcov/) and never printed is invisible to it.
The QA contract per task
- After the manual pass, this task's scenarios are added to the suite (e2e-automation-project).
- The suite — including the new tests — is green: run it yourself and/or confirm the pipeline (
get_pipeline_status) after your changes land. - Only then does the task move to
human_uat, with the suite result quoted in the evidence comment.
New scenarios accumulate: every finished task widens the regression net that all future tasks must pass.
Red Flags
- The automation project exists but no pipeline or CI job executes it.
test_commandruns unit tests only, and the e2e suite is "run manually sometimes".- The entrypoint skips the suite when dependencies are missing and still exits 0.
- Tests pass locally but the pipeline lacks the runtime/config to run them — fix the pipeline, don't drop the tests.
- The suite runs but prints no coverage total (no
-cover/--coverage/--covflag), so the pipeline's coverage badge never appears even though tests genuinely ran.
04 同仓库更多 Skills
visual-craft
暂无简介
ux-copy
Use when you write any words a user reads in a mockup or a hand-off - content language and voice, buttons, labels, errors, empty states, confirmations, numbers and dates, and the longest-string check
screen-mockup-html
Use when you draw screens as the design task's HTML mockup - ONE canvas document holding every screen and every variant side by side, the sanitizer-safe recipe, phone and wide frames, the states grid, static control states, and the loopback self-check with screenshots at 375, 768 and 1440 attached to the task
performance-awareness
How the performance score moves for your role. Use when the score context shows a declining trend, or before handing off work you have not self-checked against every criterion.
design-variants
Use when a screen design needs variants - how many, variants as side-by-side rows on the one canvas, what makes two variants genuinely different, recommending one, how the human's choice is read, and narrowing to the chosen one in revision
design-system-authoring
Use when you derive, change or extend a design system - where tokens live per stack, the DTCG token tree, role-based names for both themes, base vs layer vs drift, the DESIGN.md section order, the inventory format, the contrast check and the lint every proposal returns