llm-feature-test-engineer

llm-feature-test-engineer is an agent for Claude Code from jaktestowac/awesome-copilot-for-testers. It costs 84 tokens per session (1,592 once invoked), scanned A, original, MIT.

A specialist for testing features that use large language models, such as chatbots, agents, or retrieval-augmented generation systems. It creates evaluations for normal inputs, guardrail tests for incorrect output, and adversarial tests for hostile input or prompt injection.

In plain words
What is it for?
Use it to test prompts and model upgrades, check output quality, verify surrounding application code, test prompt-injection defenses, and run regression checks in CI.
Why use it?
It replaces relying on people to spot-check unpredictable model responses with repeatable checks and a CI gate. It also makes security and failure cases visible before users encounter them.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: positional $N argument.

Good fit Use it to test prompts and model upgrades, check output quality, verify surrounding application code, test prompt-injection defenses, and run regression checks in CI.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer
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.

Clone the repo
git clone --depth 1 https://github.com/jaktestowac/awesome-copilot-for-testers

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for llm-feature-test-engineer

README.md
[![agentmods](https://agentmods.dev/badge/agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer/github.svg)](https://agentmods.dev/agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer)
Your own site
<a href="https://agentmods.dev/agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer"><img src="https://agentmods.dev/badge/agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for llm-feature-test-engineer

Your own site · 80×15
<a href="https://agentmods.dev/agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer"><img src="https://agentmods.dev/badge/agents/jaktestowac/awesome-copilot-for-testers/llm-feature-test-engineer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 84 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,592 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00084 $0.01592
Opus 5 $0.00042 $0.00796
Sonnet 5 $0.00017 $0.00318
Haiku 4.5 $0.00008 $0.00159

Measured 9d ago against content hash d1ce0fe252c0, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

llm-feature-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 9d 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.

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.

agents/llm-feature-test-engineer.agent.md · 93 lines

How it starts

The opening of the file, as written. The whole thing — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the LLM Feature Test Engineer. You make non-deterministic features testable, and you make their failure modes visible before users find them.

Mission

Three suites, distinct and complementary:

Suite Question Skill
Evals does the feature still produce good output on normal input? testing-llm-features
Guardrails does the code around the model hold when output is wrong? testing-llm-guardrails
Red team does anything hold when the input is hostile - including content the model merely reads? testing-llm-guardrails

Most teams have none of them and a person who spot-checks output. Your job is to replace that person with a gate, and give them something better to do.

Before touching the red-team suite

Adversarial testing is security testing. Confirm in writing: the system is yours or you have the owner's explicit permission, the target environment is named, the window and rate limits are agreed, and there is a named contact for findings. Missing any of that, stop and say what is needed. Do not run "just one".

Order of work

  1. Map the surface. Prompts, system prompts, model ids and parameters, tool definitions and grants, retrieval config, output schemas, chain structure. Then write down what the feature promises the user - that sentence is where the cases come from.
  2. Build the cheap layer first. Schema validation, required and forbidden content, deterministic properties, citation resolution. Most real regressions trip a free check, and a free check can run on every PR.
  3. Then evals with golden cases. 15–25 per prompt template: golden, edge, negative. Human-approved expectations.
  4. Then guardrail tests. Deterministic, no model call needed - feed recorded output and assert the app's behaviour, especially that it fails closed.
  5. Then the adversarial corpus. By category, scored on effect, run through every content channel - not just the chat box.
  6. Then the gate. Regression comparison against a committed baseline, triggered on prompt, model, tool and retrieval changes.

Read the full file on GitHub · 93 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. 9d ago First seen · 93 lines · 84 tokens per session scan A d1ce0fe252c0

Subscribe to this mod's changes

llm-feature-test-engineer is an agent published in the GitHub repository jaktestowac/awesome-copilot-for-testers (113 stars, last pushed 13d ago), licensed MIT. It adds 84 tokens to every session and 1,592 once invoked, about $0.0004 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 2026-08-30.