laststance/qa-team

QA Agent Team

Codex Compatibility

When running this skill in Codex, translate Claude Code-only primitives before acting: AskUserQuestion -> chat/request_user_input, TodoWrite -> update_plan, Task/TaskCreate/TeamCreate/SendMessage -> spawn_agent/send_input/wait_agent when available and allowed, and EnterPlanMode/ExitPlanMode -> a concise chat plan plus explicit approval. Resolve Read/Write/Edit/Bash/WebSearch/WebFetch to Codex file/shell/web tools, and map ~/.claude/... paths to ~/.agents/... or ~/.codex/... unless the task explicitly targets Claude Code.

Cursor Compatibility

When running this skill in Cursor Agent, translate Claude Code-only primitives before acting: AskUserQuestion -> AskQuestion; TodoWrite -> Cursor TodoWrite or an equivalent checklist; Task/TaskCreate/TeamCreate/SendMessage/multi-agent flows -> Cursor Task (subagents), parallel Tasks, or run_in_background when allowed (TeamCreate/SendMessage may have no exact match); EnterPlanMode/ExitPlanMode -> Plan mode (SwitchMode / CreatePlan) plus explicit user approval. Resolve Read/Write/Edit/StrReplace/Bash/web/search/MCP via Cursor Composer or Agent equivalents. MCP names written as mcp__server__tool typically map to call_mcp_tool with configured server identifiers. Map ~/.claude/... to ~/.cursor/skills/, .cursor/skills/, and .cursor/rules/ unless the task explicitly targets Claude Code.

Comprehensive QA verification team with 5 specialist perspectives + 1 team lead.

Team Composition

| Role | Model | Agent/Skill Base | Perspective | |------|-------|-----------------|-------------| | qa-lead | opus | quality-engineer | Orchestration, aggregation, final gate | | visual-tester | opus | gui-phd-* | Layout, responsive, rendering | | functional-tester | opus | feature-validator (enhanced) | User flows + Impact Propagation | | hig-tester | opus | web-design-guidelines | Apple HIG, WCAG AA+ | | edge-case-tester | opus | quality-engineer | Long text, 100+ records, overflow | | ux-tester | opus | ph-quality-gate | Dark-on-dark, missing feedback, inconsistency |

Activation Flow

Step 1: Prerequisites

1. Ensure dev server / app is running (ask user if unclear)
2. mkdir -p claudedocs/qa/screenshots
3. Detect platform from package.json (or --platform flag):
   • Has "electron" → Electron
   • Has expo / app.json → iOS/Expo
   • Has .xcodeproj / Package.swift macOS target → macOS
   • Default → Web

Step 2: Create Agent Team

Use TeamCreate tool with team_name "qa-team" and description "QA verification team".

Do NOT do QA work yourself. Only create the team and let teammates work.

Then spawn 6 teammates using the Task tool with team_name: "qa-team":

qa-lead (subagent_type: "general-purpose"):

• Read ~/.claude/agents/quality-engineer.md
• Create task list with 8 tasks (see dependency graph below)
• Wait for all 5 reports, then aggregate into claudedocs/qa/qa-summary.md
• Apply Final Composite Gate from scoring rubric

visual-tester (subagent_type: "general-purpose"):

• Read ~/.claude/agents/gui-phd-web-electron.md (or gui-phd-mobile/gui-phd-macos per platform)
• Execute RALPH protocol screenshot verification at 95% threshold
• Report to claudedocs/qa/qa-visual-integrity.md

functional-tester (subagent_type: "general-purpose"):

• Read ~/.claude/agents/feature-validator.md
• Execute 4-Phase Impact Propagation Protocol
• Report to claudedocs/qa/qa-functional.md

hig-tester (subagent_type: "general-purpose"):

• Read CLAUDE.md Design Policies section
• Check Typography, Tap Areas, Colors, Spacing, Motion, Corner Radius
• Report to claudedocs/qa/qa-hig-compliance.md

edge-case-tester (subagent_type: "general-purpose"):

• Read ~/.claude/agents/quality-engineer.md
• Test empty/long text, 100+ records, boundary values, overflow
• Report to claudedocs/qa/qa-edge-cases.md
• blockedBy: functional-tester task

ux-tester (subagent_type: "general-purpose"):

• Apply PH Quality Gate Visual axis (V1-V5, /100)
• Report to claudedocs/qa/qa-ux-sensibility.md
• blockedBy: visual-tester task

Step 3: Tool Selection by Platform

| Platform | Primary Tool | |----------|-------------| | Web | playwright-cli (open --headed, snapshot, screenshot, click, fill) | | Electron | /qa-electron skill (playwright-cli attached to Electron CDP) | | iOS/Expo | mcp__ios-simulator__* | | macOS | mcp__mac-mcp-server__* |

Before any browser interaction (Web or Electron): invoke /dnd to load the drag-and-drop verification protocol. Ref-based drag reports false success on dnd-kit and similar libraries — load this preemptively, even when DnD is not yet known to be involved.

Step 4: Enter Delegate Mode

After spawning all teammates, tell the user to press Shift+Tab to enter delegate mode. The lead should coordinate — YOU should not write any QA reports yourself.

Step 5: Monitor & Aggregate

• qa-lead waits for all 5 reports in claudedocs/qa/
• Computes composite score using scoring rubric weights
• Generates claudedocs/qa/qa-summary.md with final verdict

Step 6: Final Quality Gate

| Component | Weight | PASS Threshold | |-----------|--------|---------------| | Visual | 25% | 95% Triple-Criteria | | Functional | 30% | 95% pass rate, P0=0 | | HIG | 15% | 80/100 composite | | Edge Cases | 15% | 0 crash | | UX Sensibility | 15% | PH Visual 75/100, critical=0 |

>= 85: PASS / 65-84: CONDITIONAL PASS / < 65: FAIL

Task Dependency Graph

Task 1: [qa-lead] Platform Detection & Test Plan
  ├── Task 2: [visual-tester]     Visual Integrity         (parallel)
  ├── Task 3: [functional-tester] Functional Correctness   (parallel)
  ├── Task 4: [hig-tester]        Apple HIG Compliance     (parallel)
  ├── Task 5: [edge-case-tester]  Edge Cases               (blockedBy: 3)
  ├── Task 6: [ux-tester]         UX Sensibility           (blockedBy: 2)
  ├── Task 7: [qa-lead]           Report Aggregation       (blockedBy: 2,3,4,5,6)
  └── Task 8: [qa-lead]           Final Quality Gate       (blockedBy: 7)

Output Structure

claudedocs/qa/
├── qa-test-plan.md              # Platform + test plan
├── qa-visual-integrity.md       # Visual report
├── qa-functional.md             # Functional report (with impact verification)
├── qa-hig-compliance.md         # HIG report
├── qa-edge-cases.md             # Edge case report
├── qa-ux-sensibility.md         # UX sensibility report
├── qa-summary.md                # Aggregated final verdict
└── screenshots/
    ├── visual_*.png
    ├── func_*.png / func_*_impact_*.png
    ├── hig_*.png
    ├── edge_*.png
    └── ux_*.png

Flags

| Flag | Effect | |------|--------| | --platform <type> | Force platform (skip auto-detection) | | --skip <perspective> | Skip one or more perspectives | | --quick | Only visual + functional (3 teammates instead of 6) |

References

• references/qa-scoring-rubric.md — Detailed scoring criteria per perspective
• workflows/web-workflow.md — Web platform test workflow
• workflows/electron-workflow.md — Electron test workflow
• workflows/mobile-workflow.md — iOS/Expo test workflow
• workflows/macos-workflow.md — macOS test workflow

Completion Criteria

QA session is complete when:

• All 5 individual reports exist with verdicts
• All reports have supporting screenshots
• Composite score calculated in qa-summary.md
• Final verdict rendered (PASS / CONDITIONAL PASS / FAIL)