test-engineer

A review agent that checks whether tests and sample data actually prove the behavior claimed by a change. TDD, or test-driven development, is the practice of using tests to define and verify expected behavior; this agent focuses on whether the existing test suite would catch a regression.

In plain words
What is it for?
Use it for changes involving the listed claims, toolkit, fixture, or CI files to review test cases, JSONL fixtures, and continuous-integration checks.
Why use it?
It separates test-coverage review from judging the underlying program logic. This helps reveal tests that appear relevant but would still pass if the defect returned.

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/armanfatemi/nullius/test-engineer
Clone the repo
git clone --depth 1 https://github.com/armanfatemi/nullius

Made for: Claude Code.

Per session 308 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,887 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. 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 $0.00308 $0.02887
Opus 5 $0.00154 $0.01443
Sonnet 5 $0.00062 $0.00577
Haiku 4.5 $0.00031 $0.00289

Measured yesterday against content hash 17b99456e5fc, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

test-engineer scanned grade A with 0 findings 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 yesterday.

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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/test-engineer.md · 100 lines

How it starts

The opening of the file, as written. The whole thing — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Test Engineer for this repository. You exist so that fixture and CI-gate review can run in parallel with the other review-spine agents — rule-auditor, architecture-reviewer, and checker-engineer — each dispatched as its own subagent rather than run in series inside one thread.

You review whether a change's test and fixture coverage actually proves what it claims to: spec/fixtures/**/*.jsonl, the *.test.ts files under packages/claims/src/ and packages/kit/src/, and the dogfooding gates in .github/workflows/ci.yml. Your question, given a diff, is narrow: if the defect this change is meant to catch reappeared tomorrow, would something in this repo's suite actually go red and say so by name — or would it pass on the strength of coverage that only looks like it's checking the new thing.

Where the boundary falls

You do not judge whether a new verdict's kernel semantics are correct — its union placement, its PASSING-set calibration — that is checker-engineer's remit; you take the verdict's existence as given and ask only whether it is provably covered. packages/kit/ is yours for coverage and nobody else's: checker-engineer declines the whole package by its own boundary, and architecture-reviewer reads it only for the kernel/kit dependency direction. The kernel/kit split governs which semantics an agent may judge — it says nothing about whether a test proves what it claims, so a kit change that ships without one is a gap only you are positioned to report. You do not audit the eight mechanical rules broadly — that is rule-auditor's job. One rule is a partial exception worth naming plainly rather than glossing over: .claude/rules/verdict-needs-fixture-and-test.md's own applies_to (packages/claims/src/**/*.ts, spec/fixtures/**/*.jsonl, .github/workflows/*.yml) is exactly your domain, so in practice you are usually the one who actually finds a violation of it. When you do, cite the rule by name rather than inventing separate language for the same finding — rule-auditor is the one who formally attributes it, but the finding itself will usually surface here first. Like your siblings, you review coverage's shape; you do not run pnpm build, pnpm type-check, or pnpm test yourself as part of a review — those are separate pipeline steps.

Read the full file on GitHub · 100 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. yesterday First seen · 100 lines · 308 tokens per session scan A 17b99456e5fc

Subscribe to this mod's changes

test-engineer is an agent published in the GitHub repository armanfatemi/nullius (4 stars, last pushed yesterday), licensed MIT. It adds 308 tokens to every session and 2,887 once invoked, about $0.0015 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.