Menú
Route natural-language QA planning, execution, tracker, and bug-draft requests through the Newton QA CLI artifact contract.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
| name | newton-qa-workflow |
| description | Route natural-language QA planning, execution, tracker, and bug-draft requests through the Newton QA CLI artifact contract. |
Use this skill when the user asks for QA planning, sprint QA, ticket QA, test scope, executable QA scenarios, web QA runs, QA evidence, tracker updates, or bug ticket drafts in a repository that can use Newton.
Newton is file-first. Prefer the newton qa ... CLI over ad-hoc prompts or one-off test notes. Preserve artifact paths in your response so another agent or CI job can replay the work.
Newton's release-quality web flow is web-first:
--target web executable scenarioUse --target web for generated scenarios and --backend playwright for execution unless the user explicitly asks for planning-only output. Mobile execution is not release-ready. If a user asks for mobile QA, keep the response to planning/risk analysis or explain that Newton's release-quality execution path is currently web-first.
newton version
bash scripts/install.sh
If the repository is not checked out, use the official installer:
curl -fsSL https://raw.githubusercontent.com/qyinm/Newton/main/scripts/install.sh | bash
Map natural-language requests to the narrowest Newton CLI flow:
newton qa plan --target web, then validate the scenario.--backend playwright, then generate a report.qa-run-tracker.md from a run directory or from a manual failure note.When the user asks for a QA plan, sprint QA, test scope, risk map, QA estimate, checklist, test cases, or automation candidates:
newton qa plan-bundle <ticket-or-context.md> \
--source <policy-or-design.md> \
--out <qa-plan-root>
newton qa bundle-validate <bundle-dir>
Use every relevant source file the user provided. Report the input files and generated artifacts:
qa-scope.mdqa-estimate.mdchecklist.mdtest-cases.csvrisk-map.mdautomation-candidates.mdqa-run-tracker.mdmanifest.jsonExample user request:
"Plan QA for this ticket, but do not run anything yet."
Example commands:
newton qa plan-bundle docs/tickets/login.md \
--source docs/policies/auth.md \
--out qa/login \
--bundle-dir-name plan
newton qa bundle-validate qa/login/plan
Final response emphasis: list the input files, planning artifacts, commands run, and say execution remains unverified.
When the user asks for an executable scenario, smoke test, or agent-runnable QA flow:
newton qa plan <ticket-or-context.md> \
--agent template \
--target web \
--out <scenario-dir>
newton qa validate <scenario.yaml>
If the user requests an external planning agent, use --agent codex or --agent claude, then preserve the prompt, raw output, accepted scenario, and provenance paths in the summary.
Example user request:
"Turn this ticket into a web smoke scenario."
Example commands:
newton qa plan docs/tickets/login.md \
--agent template \
--target web \
--base-url http://127.0.0.1:8000 \
--out qa/login/scenario
newton qa validate qa/login/scenario/login-smoke.generated.yaml
Final response emphasis: list the scenario path, validation command, provenance files when available, and remaining scenario risk.
When the user asks to run QA, verify a scenario, test a local fixture, or collect evidence:
newton qa validate <scenario.yaml>
newton qa run <scenario.yaml> \
--target <target-id> \
--backend playwright \
--base-url <url> \
--out <runs-dir>
newton qa report <runs-dir>/<run-id>
Report the run id, status, result path, report path, and evidence paths. For failures, call out screenshot and trace artifacts when present.
Example user request:
"Run the generated login scenario against staging and collect evidence."
Example commands:
newton qa validate qa/login/scenario/login-smoke.generated.yaml
newton qa run qa/login/scenario/login-smoke.generated.yaml \
--target web \
--backend playwright \
--base-url https://staging.example.com \
--out qa/login/runs
newton qa report qa/login/runs/<run-id>
Final response emphasis: list the run id, result path, report path, screenshot or trace evidence paths, commands run, and remaining risk.
When the user asks to update QA status from a run:
newton qa tracker-update-from-run <qa-run-tracker.md> \
--item <number> \
--env <dev|stg|prod> \
--run <run-dir>
If the user provides a manual failure instead of a run directory, use:
newton qa tracker-update <qa-run-tracker.md> \
--item <number> \
--env <dev|stg|prod> \
--status failed \
--notes "<actual result and evidence>"
Example user request:
"Update the tracker from the failed staging run."
Example commands:
newton qa tracker-update-from-run qa/login/plan/qa-run-tracker.md \
--item 5 \
--env stg \
--run qa/login/runs/<run-id>
Final response emphasis: list the tracker path, updated item, linked run report, evidence paths, and commands run.
When the user asks to create a bug ticket draft from QA state:
newton qa bug-draft <qa-run-tracker.md> \
--out <bug-ticket-draft.md>
Example user request:
"Draft the bug ticket from that failed checklist item."
Example commands:
newton qa bug-draft qa/login/plan/qa-run-tracker.md \
--out qa/login/bug-ticket-draft.md
Final response emphasis: list the bug draft path, failed checklist item, linked evidence, commands run, and remaining risk.
Every final response after using Newton should include this checklist:
Do not claim that QA passed unless newton qa run or the relevant validation command actually passed in the current turn. If you only generated a plan, say that execution remains unverified.