~ / rigs / huseyinozturk / agentic-qa-team-demo

huseyinozturk/agentic-qa-team-demo

Runnable demo of an agentic QA team: five Claude agents, tool allowlists, evidence-based handoffs, prompt-injection defences, on a mock tokenized-asset app

↗ GitHub ★ 0 MIT updated today project Claude Code
field: unknown workflow: Testing & TDD
share on X
ARCHETYPE
Fort Knox
Deny lists, pre-tool hooks, sandboxing. Nothing touches prod without a signature.
CONTEXT TAX · EVERY TURN
~277 tokens
Featherweight · median rig: 2.3k · breakdown
GUARDRAILS
4/5
Blocks destructive commands · Protects secrets · Pre-tool screening hook · No YOLO mode · details
Quality 63/100 Guardrails v2

Copy this rig

# review before running: this installs third-party code
$ npx degit huseyinozturk/agentic-qa-team-demo/.claude ./rig-agentic-qa-team-demo  # inspect, then merge into .claude/

MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.

$ curl -fsSL --create-dirs -o .claude/agents/analyst.md https://raw.githubusercontent.com/huseyinozturk/agentic-qa-team-demo/main/.claude/agents/analyst.md
$ curl -fsSL --create-dirs -o .claude/agents/automation-engineer.md https://raw.githubusercontent.com/huseyinozturk/agentic-qa-team-demo/main/.claude/agents/automation-engineer.md
$ curl -fsSL --create-dirs -o .claude/agents/qa-lead.md https://raw.githubusercontent.com/huseyinozturk/agentic-qa-team-demo/main/.claude/agents/qa-lead.md
$ curl -fsSL --create-dirs -o .claude/agents/runner.md https://raw.githubusercontent.com/huseyinozturk/agentic-qa-team-demo/main/.claude/agents/runner.md
$ curl -fsSL --create-dirs -o .claude/agents/watcher.md https://raw.githubusercontent.com/huseyinozturk/agentic-qa-team-demo/main/.claude/agents/watcher.md

Merge into .claude/settings.json (project) or ~/.claude/settings.json (user). Hook commands reference scripts in the source repo — copy those too.

{
  "permissions": {
    "deny": [
      "Bash(rm -rf:*)",
      "Bash(git push:*)",
      "Bash(git commit:*)",
      "Bash(git reset:*)",
      "Bash(curl:*)",
      "Bash(wget:*)",
      "Bash(sudo:*)",
      "Bash(pkill maestro:*)",
      "Read(./.env*)",
      "Read(~/.ssh/**)",
      "Edit(./.claude/**)",
      "Write(./.claude/**)"
    ]
  },
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash|Write|Edit|mcp__.*",
        "hooks": [
          {
            "type": "command",
            "command": "node \"$CLAUDE_PROJECT_DIR/.claude/hooks/pretool-guard.mjs\""
          }
        ]
      }
    ]
  }
}

Subagents (5)

analyst
model: opus
Analyst. Writes test cases from tickets and acceptance criteria with a named source for every step, and lists gaps. Use when the Lead hands over a ticket that needs a case.
automation-engineer
model: sonnet
Automation engineer. Turns ready cases into Maestro flows that pass lint and one run, with a correct header and tags. Use when the Lead hands over ready cases.
qa-lead
model: opus
QA Lead. The one agent the owner talks to. Splits requests into task files for the other roles and reviews every handoff against evidence. Use for planning, delegation and review.
runner
model: sonnet
Runner. Starts a suite when the device is free, reads the results and writes the morning report with failure classes, evidence and what changed since last night. Use for the nightly report or "run the
watcher
model: sonnet
Watcher. Reads the fetched, redacted tracker and chat data of a day and reports what changed for testing. Read-only: no Bash, no network. Use for "what changed" or the scheduled morning run.

Hooks (1)

eventmatcherruns
PreToolUseBash|Write|Edit|mcp__.*node "$CLAUDE_PROJECT_DIR/.claude/hooks/pretool-guard.mjs"

Permissions

deny (12)
Bash(rm -rf:*)
Bash(git push:*)
Bash(git commit:*)
Bash(git reset:*)
Bash(curl:*)
Bash(wget:*)
Bash(sudo:*)
Bash(pkill maestro:*)
Read(./.env*)
Read(~/.ssh/**)
Edit(./.claude/**)
Write(./.claude/**)
ask (0)
—
allow (0)
—

Similar rigs

copied ✓