Agent skill
qa
Systematically QA test a web application and fix bugs found. Three tiers: Quick (critical/high only), Standard (+ medium), Exhaustive (+ cosmetic). Produces before/after health scores, fix evidence, and ship-readiness summary. Use when testing a feature, finding bugs, or verifying deployments.
Install this agent skill to your Project
npx add-skill https://github.com/DenchHQ/DenchClaw/tree/main/skills/gstack/qa
SKILL.md
QA — Test, Fix, Verify
You are both a QA engineer AND a bug-fix engineer. Test like a real user — click everything, fill every form, check every state. When you find bugs, fix them with atomic commits, then re-verify.
Related skills: qa-only | review | ship | investigate
Setup
Parse the request for:
| Parameter | Default |
|---|---|
| Target URL | Auto-detect or required |
| Tier | Standard |
| Scope | Full app or diff-scoped |
Tiers:
- Quick: Fix critical + high severity only
- Standard: + medium severity (default)
- Exhaustive: + low/cosmetic severity
Diff-aware mode (automatic when on feature branch with no URL): Analyze git diff main...HEAD --name-only to scope testing to changed files/routes.
Phases
Phase 1: Baseline
- Navigate to the target URL
- Check if the page loads
- Check console for JavaScript errors
- Check network for failed requests
- Take a screenshot as baseline
Phase 2: Authenticate (if needed)
If auth is required, handle login flow or cookie import.
Phase 3: Exploration
Navigate every reachable page. For each page:
- Screenshot the initial state
- Console errors — any JS exceptions?
- Network errors — any failed requests?
- Interactive elements — buttons, forms, links
- Responsive — does it work on mobile (375px)?
- Empty states — what happens with no data?
- Edge cases — very long text, special characters, rapid clicking
Phase 4: Issue Documentation
For each issue found:
ISSUE #N: [Title]
Severity: CRITICAL / HIGH / MEDIUM / LOW / COSMETIC
Page: [URL]
Steps to reproduce:
1. [step]
2. [step]
Expected: [what should happen]
Actual: [what happens]
Evidence: [screenshot or console output]
Phase 5: Fix
For each issue at or above the tier threshold:
- Find the root cause in source code (use investigate methodology if complex)
- Fix it — smallest change that eliminates the problem
- Write a regression test that fails without the fix
- Atomic commit:
fix: [description of what was fixed] - Re-verify — navigate back to the page and confirm the fix
Phase 6: Report
QA REPORT
═══════════════════════════════════════
URL: [target]
Tier: [Quick/Standard/Exhaustive]
Pages tested: N
Issues found: N (X critical, Y high, Z medium)
Issues fixed: N
Tests added: N
Health score: X/100 (before) → Y/100 (after)
───────────────────────────────────────
Ship ready: YES / NO / WITH CONCERNS
═══════════════════════════════════════
Phase 7: Ship Readiness
- YES — All critical and high issues fixed. Tests pass.
- NO — Unfixed critical issues remain.
- WITH CONCERNS — Fixed what we could, but flagged issues need human judgment.
After QA, proceed to ship if ship-ready.
Important Rules
- Check for clean working tree before starting. QA needs atomic commits for each fix.
- Every fix gets a regression test. No exceptions.
- Re-verify every fix by navigating back and confirming.
- Never mark ship-ready with unfixed critical issues.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
dench-integrations
Connected app integration recipes for Dench Integrations (Gmail, Slack, GitHub, Notion, Google Calendar, Linear, Stripe, YouTube, and 500+ more)
gstack-workflow
Structured AI-assisted development workflow with specialist roles — Think, Plan, Build, Review, Test, Ship, Reflect. Adapted from garrytan/gstack (MIT).
plan-ceo-review
CEO/founder-mode plan review. Rethink the problem, find the 10-star product, challenge premises. Four modes: SCOPE EXPANSION, SELECTIVE EXPANSION, HOLD SCOPE, SCOPE REDUCTION. Use when reviewing strategy, questioning scope, or before engineering review.
investigate
Systematic root-cause debugging. Four phases: investigate, analyze, hypothesize, implement. Iron Law: no fixes without root cause investigation first. Use when debugging errors, unexpected behavior, or troubleshooting.
design-consultation
Build a complete design system from scratch. Research the landscape, propose safe choices AND creative risks, generate DESIGN.md. Use when creating a new design system or establishing design foundations for a project.
land-and-deploy
Merge the PR, wait for CI and deploy, verify production health. Takes over after ship. One command from "approved" to "verified in production."
Didn't find tool you were looking for?