QA tooling recommendation
Use this skill when qa, qa-init, or qa-strategy needs to select tools for a
specific surface. It recommends; it does not install, configure, or pretend to run
tools. Read the Tooling catalog before producing a
recommendation.
Inputs
- Observed stack and the surface or objective to test.
- Runtime and operating system, including the mobile target platform when applicable.
- A probe inventory keyed by executable or prerequisite. Every supplied probe has
state (positive, negative, or not_run), result, and evidence containing
the exact command/check and its observed output.
- CI environment, available devices/browsers, budget, data restrictions, and explicit
authorization boundaries.
An absent probe means unverified; it is never rewritten as a negative result.
Procedure
- Match the requested surface and platform against
Recommendation rules in the
catalog. Do not return platform-incompatible mobile entries.
- Prefer an existing tool. Recommend migration only when a concrete coverage,
reliability, or evidence benefit outweighs its adoption cost.
- Assign
available only when the tool probe and every required probe in the catalog
are positive. Assign missing only when the tool's own supplied probe is
negative. Assign unverified when a probe is absent, not_run, or when a required
prerequisite is negative. Preserve every supplied result and evidence; never
synthesize a command result from an absent key.
- For Web/UI exploration, distinguish
playwright-cli from Playwright Test:
playwright-cli operates an isolated interactive session using observed refs,
snapshots, actions, and screenshots; Playwright Test owns repeatable suites and CI.
Interactive CLI work does not replace a deterministic test suite.
- Before stating a current version, compatibility, cost, or license, consult the
official publisher documentation and record its URL and the verification date.
Otherwise label that claim
unverified; do not infer it from memory.
- Return all candidates, including useful missing candidates, with a usable
alternative and a context-specific reason. State limitations rather than widening
scope or authorization.
The documented detection probes for the Web/UI CLI are playwright-cli --version and,
for a project-local package, npx --no-install playwright --version; the documented
local entry is npx playwright cli. These commands were reconciled with the official
publisher documentation on 2026-09-12. If an already available CLI is explicitly
authorized for exploration, the supported isolated-session sequence is
playwright-cli -s=qa-report open, playwright-cli -s=qa-report snapshot,
playwright-cli -s=qa-report screenshot --filename=report.png, and
playwright-cli -s=qa-report close.
Output
Return one Markdown table with exactly these keys, in this order:
| tool |
purpose |
availability |
evidence |
prerequisites |
alternative |
reason |
| tool name |
bounded use |
available, missing, or unverified |
exact detection result or not probed |
platform, runtime, access, and verified cost/license facts |
feasible fallback |
why it fits this context |
Do not add hidden recommendation fields. Put official source URLs and verification
dates next to any verified current claim inside prerequisites or evidence.
Failure modes
- Missing stack, surface, platform, or constraints: return candidates as
unverified
and identify the missing input in evidence and reason.
- Positive tool probe without a positive required device, browser, driver, SDK,
service, credential, or authorization probe: preserve the supplied evidence but
report the full tool as
unverified.
- Unsupported platform: do not recommend the incompatible tool; return the compatible
alternative or explain why none was verified.
- No suitable tool: return a row with
tool set to none verified, availability
unverified, and the next safe manual or built-in alternative.
Safety
- Never install any tool automatically, never emit an installation command as an
action, and never change personal or repository configuration.
- Require explicit authorization for load, penetration testing, production access, or
external effects. Tool availability does not grant that authorization.
- For
playwright-cli, use a task-owned named session. Never reuse personal profiles,
export cookies or storage, or close sessions belonging to someone else.
- Review screenshots before attaching them. Treat snapshots as normalized text.
Traces and videos are outside the v1 evidence attachment whitelist.
- Preserve data restrictions and do not send sensitive inputs to external services.
Related skills
qa routes an explicit tooling request here only when this installed skill is a
regular local file.
qa-init inventories observed tools and limitations before planning execution.
qa-strategy consumes recommendations when selecting coverage and environments.
qa-specialist-web, qa-specialist-mobile, and qa-specialist-automation apply the
chosen tool within their narrower surface contracts.
1---2name: qa-tooling3description: Recommend context-appropriate QA tools while preserving observed availability, constraints, and safer alternatives.4metadata:5 source: https://skillmd.com/skills/pwdev-solucoes/qa-tooling6---78# QA tooling recommendation910Use this skill when `qa`, `qa-init`, or `qa-strategy` needs to select tools for a11specific surface. It recommends; it does not install, configure, or pretend to run12tools. Read the [Tooling catalog](../../references/tooling.md) before producing a13recommendation.1415## Inputs1617- Observed stack and the surface or objective to test.18- Runtime and operating system, including the mobile target platform when applicable.19- A probe inventory keyed by executable or prerequisite. Every supplied probe has20 `state` (`positive`, `negative`, or `not_run`), `result`, and `evidence` containing21 the exact command/check and its observed output.22- CI environment, available devices/browsers, budget, data restrictions, and explicit23 authorization boundaries.2425An absent probe means `unverified`; it is never rewritten as a negative result.2627## Procedure28291. Match the requested surface and platform against `Recommendation rules` in the30 catalog. Do not return platform-incompatible mobile entries.312. Prefer an existing tool. Recommend migration only when a concrete coverage,32 reliability, or evidence benefit outweighs its adoption cost.333. Assign `available` only when the tool probe and every required probe in the catalog34 are `positive`. Assign `missing` only when the tool's own supplied probe is35 `negative`. Assign `unverified` when a probe is absent, `not_run`, or when a required36 prerequisite is negative. Preserve every supplied `result` and `evidence`; never37 synthesize a command result from an absent key.384. For Web/UI exploration, distinguish `playwright-cli` from Playwright Test:39 `playwright-cli` operates an isolated interactive session using observed refs,40 snapshots, actions, and screenshots; Playwright Test owns repeatable suites and CI.41 Interactive CLI work does not replace a deterministic test suite.425. Before stating a current version, compatibility, cost, or license, consult the43 official publisher documentation and record its URL and the verification date.44 Otherwise label that claim `unverified`; do not infer it from memory.456. Return all candidates, including useful missing candidates, with a usable46 alternative and a context-specific reason. State limitations rather than widening47 scope or authorization.4849The documented detection probes for the Web/UI CLI are `playwright-cli --version` and,50for a project-local package, `npx --no-install playwright --version`; the documented51local entry is `npx playwright cli`. These commands were reconciled with the official52publisher documentation on 2026-09-12. If an already available CLI is explicitly53authorized for exploration, the supported isolated-session sequence is54`playwright-cli -s=qa-report open`, `playwright-cli -s=qa-report snapshot`,55`playwright-cli -s=qa-report screenshot --filename=report.png`, and56`playwright-cli -s=qa-report close`.5758## Output5960Return one Markdown table with exactly these keys, in this order:6162| tool | purpose | availability | evidence | prerequisites | alternative | reason |63|---|---|---|---|---|---|---|64| tool name | bounded use | available, missing, or unverified | exact detection result or `not probed` | platform, runtime, access, and verified cost/license facts | feasible fallback | why it fits this context |6566Do not add hidden recommendation fields. Put official source URLs and verification67dates next to any verified current claim inside `prerequisites` or `evidence`.6869## Failure modes7071- Missing stack, surface, platform, or constraints: return candidates as `unverified`72 and identify the missing input in `evidence` and `reason`.73- Positive tool probe without a positive required device, browser, driver, SDK,74 service, credential, or authorization probe: preserve the supplied evidence but75 report the full tool as `unverified`.76- Unsupported platform: do not recommend the incompatible tool; return the compatible77 alternative or explain why none was verified.78- No suitable tool: return a row with `tool` set to `none verified`, availability79 `unverified`, and the next safe manual or built-in alternative.8081## Safety8283- Never install any tool automatically, never emit an installation command as an84 action, and never change personal or repository configuration.85- Require explicit authorization for load, penetration testing, production access, or86 external effects. Tool availability does not grant that authorization.87- For `playwright-cli`, use a task-owned named session. Never reuse personal profiles,88 export cookies or storage, or close sessions belonging to someone else.89- Review screenshots before attaching them. Treat snapshots as normalized text.90 Traces and videos are outside the v1 evidence attachment whitelist.91- Preserve data restrictions and do not send sensitive inputs to external services.9293## Related skills9495- `qa` routes an explicit tooling request here only when this installed skill is a96 regular local file.97- `qa-init` inventories observed tools and limitations before planning execution.98- `qa-strategy` consumes recommendations when selecting coverage and environments.99- `qa-specialist-web`, `qa-specialist-mobile`, and `qa-specialist-automation` apply the100 chosen tool within their narrower surface contracts.