agentleFS
Sn4kyGitSn4kyGit/qa-agent-blueprint/.claude/skills/test-ticket/SKILL.md

test-ticket

Skill2 starsChanged 7 days agoApache-2.0View on GitHub

Test a JIRA ticket end to end — read the ticket, route to the project, verify/deploy stage, execute the test strategy, produce evidence, report the verdict to JIRA. Use for "/test-ticket <KEY-123>", "teste <KEY-123>", "prüf das Ticket", or any verify-a-ticket request. Not for re-testing fixed bugs (retest-bug) or writing repeatable specs (test-automation).

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Add a product only when ## Products exists and the registry matches, using the same targeted Learnings lookup …

    line 32
  2. Very specific:

    Seal the run context once: Write evidence/<KEY>/<TICKET>/<RUN>/context-capsule.json per docs/pipeline/context-capsule.md, embedding the …

    line 52
  3. Very specific:

    Verify execution_summary.json and the required evidence, then report through report-to-jira on PASS/FAIL/BLOCKED/NEEDS_REVIEW exactly as …

    line 92

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/test-ticket — dispatch the full ticket workflow

The primary entry point. The agent loading this skill is the main orchestrator: it performs preflight, intake, routing and readiness exactly once. A simple run then goes directly to one execution leaf; only a complex run delegates its remaining orchestration to qa-manager.

Steps

  1. Parse the ticket key from the argument (SHOP-1234). If missing, ask for one. An explicit product (--product=ADMIN, or "test this in the admin") overrides automatic product routing. Read enough of projects/<KEY>/PROJECT.md to run the preflight and select the JIRA access channel. Unknown project prefix or missing project file → stop and run onboard-project (or tell the user to). A half-registered product → stop and run onboard-product; never guess a process.
  2. Run preflight once, before any browser or JIRA call (the blueprint's preflight.sh <KEY>; the pre-tool guard enforces it). Then make the trivial Atlassian read on mcp access. Do not navigate Playwright to stage for liveness yet. JIRA access browser is the explicit exception: intake through the playwright-driven JIRA UI per docs/rules/testing/jira-browser.md is allowed now; this exception covers JIRA only, never stage.
  3. Intake and route once. Fetch the targeted ticket fields, number the ACs, assess testability and risk, and write intake.json per docs/pipeline/01-intake.md. Route to projects/<KEY>/PROJECT.md; search its learnings.md only for the ticket's area, component, flow and failure keywords. Add a product only when ## Products exists and the registry matches, using the same targeted Learnings lookup (.claude/rules/core/product-layer.md). Load a flow map only when the mode needs it for recon, planning, retest or a charter, and treat only confirmed Areas as binding. Resolve ambiguity before stage; headless reports BLOCKED. Carry the resolved scope into every later delegation.
  4. Gate on readiness before the first stage action. Apply docs/pipeline/01a-readiness.md and write readiness.json: allowed JIRA status, no unresolved blockers, linked MR/PR merged when the project's merge policy requires it, and expected build deployed (or the explicitly configured Build verification: unavailable, reported as unverified). Prefer API/CLI checks. Only after a ready result may one Playwright navigation check stage liveness and recon or test execution begin. A missing merge/required build is BLOCKED or a deploy-to-stage handoff followed by re-verification — never a test against a known-wrong state.
  5. Read shadow mode and decide strategy from the effective Execution mode. Shadow mode, evidence and reporting remain unchanged. mcp-only uses test-executor; cli-first uses the existing suite through test-automator; hybrid requires recon before specs, the red check for new specs, and its configured live verification pass. Escalation triggers still override every mode (docs/pipeline/00-overview.md); a red test alone is not one. A routed product's feature/test-method matrix and any required specialized skill remain binding.
  6. Seal the run context once. Write evidence/<KEY>/<TICKET>/<RUN>/context-capsule.json per docs/pipeline/context-capsule.md, embedding the exact intake/readiness and the resolved effective config, relevant confirmed knowledge and source hashes. Run context-capsule.py seal and one full validate with --qa-home and --run-dir. Revision 1 is context-capsule.json; revision N>1 must be context-capsule.vN.json in the same run dir and every predecessor stays intact. The capsule is immutable; pass the selected revision path, never its full text, in a Task prompt. A later build/readiness or effective-context change returns to this orchestrator for readiness re-verification and a new revision, never recipient-side repair. Changed same/older-revision payloads conflict; identical reseals are idempotent and invalid existing hashes are never overwritten. 6b. Write testplan.md before execution starts (docs/pipeline/03-planning.md). The plan's producer is the planning agent: the main orchestrator on the Fast Path, qa-manager on the complex path (its own flow writes the file; the complex handoff may carry a plan already produced).
  7. Dispatch exactly one way:
  • Fast Path — only when intake is unambiguously testable, risk is low, no escalation trigger applies, no specialized/multi-area orchestration is needed, and the effective mode needs exactly one execution branch. The main orchestrator delegates directly to test-executor-cheap or test-automator-cheap, with the selected sealed revision path and charter/plan; no PROJECT/PRODUCT full reads, re-routing, mode/suite lookup or repeated Learnings merge. No qa-manager spawn. For hybrid, a required live verification pass is a second branch and therefore excludes the Fast Path; recon and plan approval keep their normal mode rules.
  • Complex path — anything else: invoke qa-manager with the ticket key, run dir, selected sealed revision path and any recon/plan already produced. It consumes the capsule and must not read the ticket or PROJECT/PRODUCT files, route, repeat readiness, or redo mode/suite lookup and Learnings merging. It owns the remaining mode-conformant phases and execution delegations. Announce and relay every delegation as qa-manager's "Delegation is visible" section prescribes; browser-driving leaves remain sequential.
  1. On FAIL, the orchestrator that owns the run starts failure-analyst with all failing cases, the failure packet and selected sealed revision path — the main orchestrator on the Fast Path, qa-manager on the complex path. Never make test-executor or test-automator delegate to the analyst. The same orchestrator writes/merges the analysis and applies the class-specific next step (bug-reporter, BLOCKED, NEEDS_REVIEW or reliability handling) before reporting.
  2. Verify execution_summary.json and the required evidence, then report through report-to-jira on PASS/FAIL/BLOCKED/NEEDS_REVIEW exactly as before, including the mode and shadow behavior. The summary's skills_used names every skill this run actually invoked (this one, recon, the execution leaf's skill, failure-analysis, …) — never merely planned ones (.claude/rules/core/evidence-policy.md). Finish with retro-learn capture and the mandatory learnings declaration; unexpected routing corrections follow the standing rule.

Output to the user

A short summary: ticket, project (and product, if one was routed — plus how it was resolved), verdict, JIRA comment link (or bug ticket key if FAIL) — and the key evidence files each as an absolute path, one per line (screenshot, trace, execution summary). Absolute paths are clickable in the harness UI; a relative evidence/... path resolves against an arbitrary cwd and leads nowhere. Do not restate the full test transcript — that lives in the run dir and in the JIRA comment itself (which, per docs/pipeline/06-reporting.md → "Two output channels", never contains these local paths).

Then, on an interactive macOS session, open the run dir in Finder: open "<absolute run-dir path>" — one window with every screenshot/log, Space-key preview. Skip silently on Linux, CI and headless runs; a failed open never blocks the report. Full rule: docs/pipeline/06-reporting.md → "Finder auto-open".

Hard rules that apply here (see the router)

Evidence is mandatory (the evidence gate enforces it), no secrets anywhere in the output, no full-repo scans — start from memory/INDEX.md and the project file. No stage browser navigation, recon or execution before readiness; no universal MR/PR requirement beyond the effective project's merge policy.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.