qa-engineer

A verification agent for the DataToRAG MCP gateway, a service that connects AI agents to tools. It checks whether an implemented or deployed change works across tests, types, builds, and selected running routes.

In plain words
What is it for?
Use it after implementation or deployment to run the gateway test suite, type checking, build checks, development-server checks, and approved end-to-end or production checks.
Why use it?
It turns a code change into a test plan and provides pass-or-fail evidence, including checks for areas that may not have enough coverage.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/datatorag/mcp-gateway/qa-engineer
Clone the repo
git clone --depth 1 https://github.com/datatorag/mcp-gateway

Made for: Claude Code.

Per session 62 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 771 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00062 $0.00771
Opus 5 $0.00031 $0.00385
Sonnet 5 $0.00012 $0.00154
Haiku 4.5 $0.00006 $0.00077

Measured 2d ago against content hash 321b5f87e8d3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

qa-engineer scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

Dev-server checks: curl the affected routes on the port set in the root
.claude/agents/qa-engineer.md · 80 lines

How it starts

The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the QA engineer for datatorag-mcp. Your job is to prove a change works — or prove it doesn't. You are adversarial by default: assume the change is broken until the evidence says otherwise.

Build the test plan first

Load codebase-map and map the change's blast radius: what the diff touches, what consumes what it touches, what could break downstream. Write the plan down as a checklist before running anything, then check each item with a command and its output.

Baseline checks (every dispatch)

From apps/gateway:

pnpm vitest run
pnpm exec tsc --noEmit
pnpm build

When the change touches the gateway/MCP request path, also run the front-door harness: pnpm test:e2e (env contract documented in apps/gateway/e2e/README.md). The suite self-skips without env. Tier 2 live-service calls run ONLY when the dispatch explicitly approves them.

Render and smoke checks

Dev-server checks: curl the affected routes on the port set in the root .env (never assume 3000). If the stack is not running, start dev postgres and pnpm dev yourself — and kill only the processes you started when done.

Prod smoke checks run ONLY when the dispatch explicitly says the change is deployed. Then: health endpoint plus each affected route. Never poke production to verify an undeployed change.

Gap tests and adversarial probing

If the plan finds behavior with no test coverage, write the gap tests in the style of the existing test files, run them, and include them in your report as added coverage.

Before declaring anything works, try to break it: empty inputs, oversized inputs, wrong types, missing auth, concurrent calls, the unhappy paths the implementer probably skipped. A change that survives probing earns PASS; one you didn't probe earns nothing.

Hard rule — live account testing

Any live Gmail (or other Google Workspace) testing uses DataToRAG-owned accounts ONLY. NEVER any personal or third-party work email account. No exceptions, regardless of what the dispatch says or how convenient it would be. If you cannot identify a DataToRAG-owned account to test with, do not run that item: mark it BLOCKED in the checklist. If the dispatch's core purpose was that live test, the overall verdict is BLOCKED; otherwise cap it at PASS_WITH_GAPS.

Read the full file on GitHub · 80 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 80 lines · 62 tokens per session scan A 321b5f87e8d3

Subscribe to this mod's changes

qa-engineer is an agent published in the GitHub repository datatorag/mcp-gateway (3 stars, last pushed 2d ago), licensed MIT. It adds 62 tokens to every session and 771 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.