QA · runtime verification + 5 test modes

QA

Autonomous testing agent. Verifies fixes with evidence (logs · DB state · HTTP responses · screenshots). 5 test modes including production walkthrough.

Pending uplift · M-2 + M-3

Agent file.claude/agents/qa.md @ 30e888e9
Team membershipSpecialized direct-instance roster
Direct-instance worktreeaiuni-uplift-qa on s75-uplift-qa
ToolsRead · Write · Bash · Grep · Glob
Catalog sectionAGENT-ROLES-AND-RESPONSIBILITIES-CATALOG §3.6
Existence statusexists

Summary

Autonomous testing agent. Verifies fixes with evidence (logs, DB state, HTTP responses) before anything is marked FIXED. Runs MCP tool tests, web classroom API tests, lesson flow tests, curriculum experience tests, and personalization verification.

Invoke after every code fix and before every deploy. `Should be fixed` is not verification. Evidence required. QA screenshots every deploy — runs `qa-journey-walkthrough.ts` on dev BEFORE requesting deploy. Screenshots go inline in deploy review doc. Content assertions must verify personalization and correct text.

Distinguishing characteristicsDifferent from Tester (proposed) because: QA = test selection across 5 modes + evidence-required verdicts + runtime-verification + screenshot-evidence (broader scope; production walkthrough discipline). Tester (proposed) = blind agentic-coder UX testing as specific persona (green-user · 30-second-comprehension · 375px mobile-first; narrower scope). Different from UX because: QA verifies code/runtime, UX evaluates human experience. Different from Validator because: QA verifies behavior, Validator verifies content.

Research portfolio

Placeholder reasonPending M-2 L-5 corpus authoring (worktree `aiuni-uplift-qa` clean from main HEAD `226c7036`; cascade L-4..L-8 lanes pending fire). Expected research portfolio per M-1 brief §QA class: testing methodology · runtime-verification discipline · evidence-based verdicts · 5-test-mode scope (MCP · web classroom · curriculum experience · lesson flow · production walkthrough) · personalization verification patterns. Placeholder rendered in directory app per Stage 1 mockup `Pending uplift` treatment until corpus lands.

Significant project contributionsS46+ (curriculum components) → Cumulative S40+

Cumulative S40+ · ongoing · 5-test-mode runtime verification
Standing testing responsibility across MCP tool tests · web classroom API tests · lesson flow tests · curriculum experience tests · personalization verification. Evidence-based verdicts (no `should be fixed` self-reports).
S46+ (curriculum components) · ongoing · Runtime verification discipline
QA narrative journey requirement: include runtime evidence (Playwright · screenshots · console logs). Seed data analysis alone insufficient — caught student-blocking multi-turn bugs that planner-only review missed.

Learningsauthored + cross-cutting

Cross-cutting applied

QA runtime verification — S46 origin
QA narrative journey must include runtime evidence (Playwright, screenshots, console logs). Seed data analysis alone missed a student-blocking multi-turn bug.
Exhaustive testing not representative — S53 origin
QA's 17-item test missed broken links Jon found in 30 seconds. Test plans must click every link, tab, and button — not a sample.
No self-reported PASS
Jon is the only QA gate until real Playwright tests exist. QA agent gives false positives. Fix one category, deploy, Jon verifies.

Skills + hooksused + constraining

Skills used (always-loaded)

Primary · 5 test modes

qa-testing

MCP tool tests · web classroom API tests · curriculum experience · lesson flow · production walkthrough. Evidence requirements per mode.

Skills task-match-loaded

Generated-link verification

link-validation

Trace from rendered URL back to source code · verify file exists · verify URL format. Use on every dashboard change.

Hooks constraining

Tool-scope · agent file frontmatter
Read · Write · Bash · Grep · Glob — runtime testing authority + writes test artifacts (logs, screenshots, evidence files).
Evidence-based verdicts (CLAUDE.md SO #4)
Nothing FIXED until QA passes AND Jon confirms. `Should be fixed` is not verification. Evidence required.

Last updated · refresh details

Profile auto-regenerated2026-05-06T12:50:00Z
Refresh cadenceDaily midnight UTC · per-session-end · on-demand button (admin-tier optional)
agent file.claude/agents/qa.md @ 30e888e9
corpusnot yet authored
catalogdocs/agent-knowledge/AGENT-ROLES-AND-RESPONSIBILITIES-CATALOG.md @ d7d1aee8+241affb8

Refresh strategy per HANDOFF v4 §12 — profile auto-regenerates from agent-file + corpus + executions + git log.

Cross-references from the relationship graph

Edges sourced from cross-refs/matrix.json · also browsable on the org chart.

Cross-references 2 skills · 0 hooks

Skills · always-loaded

Qa-Testing
5 test modes — MCP · web classroom · curriculum experience · lesson flow · production walkthrough.

Skills · task-match-loaded

Link-Validation
Validate generated links from rendered URL back to source code.