qqa

name	Qqa
description	Unified QA entry point for planning test documentation, running single-pass scenarios, and orchestrating the multi-agent QA council.
user_invocable	true
metadata	{"author":"inho","version":"1.0.0"}
invocation_trigger	When the user asks for QA planning, scenario/E2E execution, or the QA council.
recommendedModel	sonnet

Qqa

Unified QA entry point. It preserves the behavior of the existing QA skills while exposing a single command surface:

Qqa <plan|run|council> [args]

The original skills remain valid and coexist:

• Qqa-test-planner
• Qscenario-test
• Qqa-council

Subcommands

Subcommand	Reference	Summary
plan	reference/plan.md	Generate test plans, manual test cases, regression suites, Figma verification notes, and bug report templates. Use this for QA documentation without executing the target.
run	reference/run.md	Generate scenario specs from a TASK_REQUEST, code path, URL, or description; execute browser/API/CLI scenarios once; verify and report results. Browser QA loop (Playwright) is live via scripts/lib/browser-driver.mjs (Phase 6): --browser drives Playwright on demand (optional dependency; detectWebProject recommends it; regression specs named qa-regression-<slug>.spec.ts).
council	reference/council.md	Run the role-separated QA council: Planner -> Explorer -> Generator -> Healer -> Reporter, with bounded agent/tool contracts and optional PR automation.

Routing

Qqa plan [target]

Use when the requested output is QA documentation:

• Test plan
• Manual test cases
• Regression suite
• Figma design verification checklist
• Bug report template

Load and follow reference/plan.md. Do not execute the system under test unless the user switches to Qqa run or Qqa council.

Qqa run [target|--rerun UUID] [--dry] [--browser|--api|--cli]

Use when the request is scenario testing, E2E scenario execution, user-flow verification, or acceptance testing. Load and follow reference/run.md.

This subcommand preserves Qscenario-test behavior: generate SCENARIO_SPEC, generate SCENARIO_CHECKLIST, execute scenarios once in browser/API/CLI mode, and write a SCENARIO_REPORT.

Qqa council [url] [mode]

Use when the request is a role-separated multi-agent QA loop over a live web app, exploratory QA plus regression codification, or a PR-triggered QA bot. Load and follow reference/council.md.

The council must preserve bounded roles:

• Planner invokes Qqa plan or Qqa run logic as a skill-backed planning step.
• Explorer is Eqa-explorer, browser-only, no source.
• Generator uses Qplaywright-expert.
• Healer is Eqa-orchestrator.
• Reporter is a mode within Eqa-orchestrator.

Never pass a Q* skill name as an Agent subagent_type; Q* entries are skills, E* entries are agents.

Selection Rules

User intent	Choose
"Create a test plan", "generate test cases", "regression suite", "bug report template"	Qqa plan
"Run scenarios", "E2E", "acceptance test", "verify this flow once"	Qqa run
"QA council", "AI QA", "exploratory + regression", "PR QA bot"	Qqa council

Compatibility

• Keep using the original skills when directly invoked.
• Do not delete or shim Qqa-test-planner, Qscenario-test, or Qqa-council in this phase.
• Qqa run is single-pass scenario execution. Long-lived browser QA loops belong to the later Phase 6 extension.