ci-suite-runner
Skill1 starsChanged yesterdayMITView on GitHub
Run a catalogued test suite in CI (GitHub Actions workflow_dispatch or an Azure DevOps pipeline run) on a chosen environment with chosen test tags, then poll it to completion and collect the results and artifacts - JUnit totals and failed tests, artifacts, failed jobs' logs - into one record. Every trigger is confirmed unless a pre-authorisation covers that suite and environment and hasn't expired; protected environments such as production always need a confirmation. Triggers once per request, never re-triggers on a timeout, accepts only the catalogue's suites, environments and tags, and hands failures to the CI failure triager.
- Reads credentials
In this repository33 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- api-contract-testerSkill
- bug-reproducerSkill
- ci-failure-triagerSkill
- code-security-auditorSkill
- cross-browser-checkerSkill
- db-migration-reviewerSkill
- dead-test-finderSkill
- exploratory-charter-writerSkill
- flaky-test-hunterSkill
- hypercare-anomaly-watcherSkill
- i18n-checkerSkill
- journey-mapperSkill
- journey-to-testsSkill
- mobile-app-auditorSkill
- mutation-testerSkill
- performance-budget-auditorSkill
- post-deploy-validatorSkill
- privacy-flow-auditorSkill
- qa-orchestratorSkill
- release-readiness-gateSkill
- requirements-traceability-auditorSkill
- resilience-proberSkill
- risk-based-test-plannerSkill
- test-case-designerSkill
- test-data-auditorSkill
- test-impact-selectorSkill
- test-suite-evaluatorSkill
- ticket-readiness-reviewerSkill
- ticket-to-tddSkill
- ui-crash-hunterSkill
- visual-regression-reviewerSkill
- wcag-auditorSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 11
Start a named suite in CI, on an environment, with a tag expression, and bring back what happened: the run link, the status as it changed, …
- Very specific: line 29
Results come from JUnit: no JUnit artifact is not_collected, never "passed".
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CI suite runner
Mission
Start a named suite in CI, on an environment, with a tag expression, and bring back what happened: the run link, the status as it changed, the conclusion, the JUnit totals and failed tests, the artifacts, and failed jobs' logs. ci-failure-triager takes a failed run from here; test-impact-selector can choose the tags.
Non-negotiable rules
- No trigger without authority. Either a grant in
preauth.jsonthat names the suite and the environment and hasn't expired, or a confirmation of this exact request (who, how, when) passed as--confirmed-by. Without one, show the exact request and stop. Never treat silence as yes. A grant's expiry is judged on today's date;--as-of(recorded in the authorisation) judges it on another date and is accepted only for a loopback CI stub, so an example can be regenerated after its grant expires. - Protected environments (production by default) always need a confirmation for the run; a pre-authorisation never covers them, even if a grant lists them.
- Only the catalogue: the suite, environment and tag markers must be in
suites.json. Anything else is refused. - Trigger once. A record that was triggered is never triggered again; a timeout is
timed_outwith the link, andpollresumes later. - Results come from JUnit: no JUnit artifact is
not_collected, never "passed"; the totals are counted from the files. - Tokens stay out of files (
GITHUB_TOKEN/GH_TOKEN,AZURE_DEVOPS_EXT_PAT); secret inputs are redacted. Tokens are never sent to another origin: artifact and log downloads redirect to storage hosts, andsafe_http.pydrops credential headers on any cross-origin redirect. Job and artifact names from the CI become plain file names (safe_name) and every write is checked to stay inside the run folder. - No "all green": say "N of M tests passed in run R".
- Never approve a pipeline's environment gate or a deployment on someone's behalf.
Inputs
| Input | Default | Notes |
catalogue | suites.json | schemas/suites.schema.json: per suite the system, pipeline, input names, environments, protected environments, markers, JUnit artifact names |
suite, environment | required | |
tags | the suite's default | Markers with and/or/not and parentheses |
ref | the suite's default | |
preauth | none | schemas/preauth.schema.json |
confirmed_by | none | From the user, for this request |
Workflow
python scripts/run_suite.py plan --catalogue suites.json --suite checkout-regression --environment staging --tags "checkout and not slow" --out run/
python scripts/run_suite.py trigger --run run/ --preauth preauth.json # or --confirmed-by "who, how, when"
python scripts/run_suite.py poll --run run/
python scripts/run_suite.py collect --run run/
python scripts/validate_suite_run.py run/suite_run.json --preauth preauth.json
python scripts/render_report.py run/suite_run.json
python scripts/self_test.py # the example stays regenerable on any date; the --as-of rules- Plan and show the user the request when no grant covers it; ask for the confirmation, and pass their
words as
--confirmed-by. - Trigger, poll (backoff, timeout), collect.
- Write the headline (totals and the conclusion), the next steps (failed tests to ci-failure-triager; what the tags excluded) and the limitations; validate and render.
Rating
See RUBRIC.md: authority, outcomes and results.
Output (schemas/suite_run.schema.json)
request, plan (the exact request), authorisations (each decision: refused with why, or sent with the
grant or the confirmation), trigger, run, polling, outcome, results, artifacts, logs, jobs,
summary, next_steps, limitations.
Edge cases
- No
workflow_dispatchon the workflow, or no parameters on the pipeline: refused; say what to add. - Approval gates: the run waits; the record says so and who must approve.
- Concurrency cancellations: recorded as the conclusion; not retried.
- A dispatch with no new run within 20 seconds:
not_started, with the reason; check the workflow's triggers and the ref.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

