菜单
Use this when QA is piloting product goals, golden paths, simulator flows, or workflow-scale usability for a real user persona.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use for screenshot-driven Flutter design work that must prove visual quality through Requirements-native claims, sanctioned Design Lab capture, and anti-slop evidence. [skill-hash:7b1c9e3]
Use Robdex communication via `robdex`, including role-scoped Requirements workflows. Role behavior lives in the base instructions. [skill-hash:8f2c6a9]
Use this when a designer needs to pilot an ad-hoc iOS simulator directly, launch Flutter from their own worktree in tmux, or trigger hot reload without the broker/device-harness path.
Script-first worktree/branch/PR workflow using local scripts under `scripts/` plus direct `git`/`gh`. Use this when you need the sanctioned mutating git and GitHub workflow tools. [skill-hash:7b3f0c2]
Ran into a sandbox issue, approval request, or privileged-exec rejection? Use this skill immediately for the sanctioned resolution path. [skill-hash:b8f4a20]
Author reliable Flutter integration tests that run with `flutter test -d flutter-tester`, especially in apps using GetX navigation and rinf-style async bridges. Use this when creating or editing `integration_test/*.dart` and avoiding deadlocks matters. [skill-hash:9c7d2f1]
| name | product-usability-qa |
| description | Use this when QA is piloting product goals, golden paths, simulator flows, or workflow-scale usability for a real user persona. |
Completion is not enough.
A story passes only when a real target user can complete it with reasonable effort, obvious navigation, clear state, and no inappropriate developer knowledge.
Simple tasks: 3-5 expected, >7 usability bug, 10+ severe usability.
Medium tasks: 6-12 expected, >15 usability bug, 20+ severe usability.
Complex tasks: 15-30 allowed only when guided, previewable where applicable, recoverable, and appropriate for advanced configuration.
AI assist is appropriate for advanced configuration, rules, workflow design, package/interface generation, marketing copy, complex automation, and import/mapping.
AI assist is not acceptable for basic operational tasks such as creating records, adding contacts, finding records, editing basic info, assigning people, sending items, approving or rejecting items, or marking work complete.
If a simple task appears to require AI assistance, report Severe Usability.
Use assets/product-usability-pilot-report.md when a full report artifact is requested.
Every pilot report must include:
Use references/product-usability-task-classes.md when task class or AI-assistance suitability is unclear.
Score each 0-3:
0-4 acceptable. 5-8 needs polish. 9-14 usability bug cluster. 15+ severe usability failure.
Do not say "passed" merely because the task eventually succeeded.
If the task took unreasonable effort, report PASS WITH USABILITY BUGS or FAIL - SEVERE USABILITY.
