メニュー
Use for the L2+ browser-QA gate. Real-browser mode (Playwright Chromium) is opt-in (--browse / SGC_QA_REAL=1); by default runs a stub returning concern (never rubber-stamps). Writes verdict + findings to reviews/{task}/qa/.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use after implementation to run independent code review - dispatches reviewer agents based on task level, ensures author-reviewer separation
Use when starting any conversation - establishes SGC commands, routes tasks to appropriate skills, and enforces system invariants
Use to capture knowledge from solved problems - extracts structured solution documents via 4 specialized agents with dedup enforcement
Use when requirements are unclear, before planning - clarifies goals, constraints, and acceptance criteria through structured questioning
Use when starting any non-trivial task - classifies task level (L0-L3), runs appropriate planning agents, produces intent document and execution plan
Use when ready to release - verifies all evidence, runs ship gate, handles deployment, triggers compound janitor
| name | qa |
| description | Use for the L2+ browser-QA gate. Real-browser mode (Playwright Chromium) is opt-in (--browse / SGC_QA_REAL=1); by default runs a stub returning concern (never rubber-stamps). Writes verdict + findings to reviews/{task}/qa/. |
Spawn qa.browser to run the QA gate; with the real-browser opt-in (--browse / SGC_QA_REAL=1) it drives a Playwright Chromium and writes verdict + screenshot refs to reviews/{task_id}/qa/qa.browser.md, flipping hasQaEvidence true to unblock the L2+ ship gate.
Real-browser mode is opt-in (Playwright). By default
qa.browserruns a stub that returnsconcern(neverpass), so the gate is never silently rubber-stamped. Enable it with--browseorSGC_QA_REAL=1— it launches a Playwright Chromium (goto→ console/page errors → screenshot → verdict). Needs a browser:npx playwright install chromium, orSGC_QA_BROWSER=chromefor system Chrome.
Core principle: "looks correct" is not evidence — with --browse / SGC_QA_REAL=1, open the browser (Playwright), run the flow, capture the proof; by default qa.browser returns concern rather than rubber-stamping.
/qa <target> --flows a,b,c/work| Directory | Access |
|---|---|
| decisions | R |
| progress | R |
| solutions | FORBIDDEN (§1) |
| reviews | W |
Plus exec:browser for the headless chromium launch.
src/commands/qa.ts (runQa)src/dispatcher/agents/qa-browser.ts — injectable browseRunner for hermetic testssrc/dispatcher/agents/playwright-runner.ts — Playwright Chromium smoke (opt-in via --browse / SGC_QA_REAL=1)hasQaEvidence in src/dispatcher/state.tsWhen this skill is invoked, dispatch to the sgc CLI:
bun src/sgc.ts qa $ARGUMENTS
qa.browser fails the verdict on pageerror / console.error captured via Playwright (page.on("pageerror") / page.on("console")). Screenshot-capture misses are surfaced as a note but don't fail the verdict.
If the chromium sandbox is broken (Ubuntu 23.10+ AppArmor user-namespace restriction, RHEL SELinux), the Playwright launch uses chromiumSandbox: false. Hermetic tests inject a fake launch/browseRunner instead of a real browser — see tests/dispatcher/playwright-runner.test.ts; the gated real-browser test is tests/eval/qa-browse-real.test.ts.
For rich interactive browser testing beyond sgc's headless qa.browser:
gs:/browse — full headless browser with navigation, screenshots, and element interaction