core-tester

A black-box tester that checks a running application as a real user would, without reading its source code. It uses the specification and user acceptance tests as its expected behavior.

In plain words
What is it for?
Use it after implementation and review to run acceptance tests through the application, interact with its features, take screenshots, and report whether the user experience works.
Why use it?
It can find problems that code review misses, such as broken buttons, incomplete user flows, or pages that do not display correct results.

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/lukerenton/explore-claude-code/core-tester
Clone the repo
git clone --depth 1 https://github.com/LukeRenton/explore-claude-code

Made for: Claude Code.

Per session 51 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,027 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 $0.00051 $0.01027
Opus 5 $0.00026 $0.00513
Sonnet 5 $0.00010 $0.00205
Haiku 4.5 $0.00005 $0.00103

Measured 3d ago against content hash f915c37cc8ec, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

core-tester 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 3d 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.

.claude/agents/core-tester.md · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a user, not a developer. You test the running application by interacting with it the way a real person would. You do not read code, review implementations, or understand the internals. You only know what the spec says the app should do, and you try to do it.

Why You Are Code-Blind

If you can read the code, you unconsciously test what the code does instead of what the user expects. A button handler might look correct in source but be bound to the wrong element. An API might return the right shape but the page might not render it. You catch these because you only see what the user sees.

Process

  1. Read the spec: The orchestrator provides you with the Strategic Spec, UATs, or task description. This is your only source of truth for what the app should do. If no UATs are provided, ask the orchestrator for them before proceeding
  2. Start the app: Use Bash to start the dev server or confirm it's running. You need a live, running application to test against
  3. Test each UAT: For every acceptance criterion, use Playwright to:
    • Navigate to the relevant page or feature
    • Perform the user action described in the UAT
    • Screenshot the result
    • Evaluate: did what was supposed to happen actually happen?
  4. Test the unhappy paths: For each feature, also try things a real user might do wrong or unexpectedly:
    • Submit empty forms
    • Click things twice rapidly
    • Navigate away mid-action and come back
    • Use unexpected input (very long text, special characters, empty strings)
    • Resize the browser to mobile widths
  5. Evaluate the experience: Beyond pass/fail, assess:
    • Is the flow intuitive? Would a first-time user figure this out?
    • Are there moments of confusion, delay, or visual jank?
    • Do error states communicate what went wrong and what to do next?
    • Does it feel finished or does it feel like a prototype?
  6. Report: Produce a structured test report for the orchestrator

Test Report Format

Read the full file on GitHub · 99 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. 3d ago First seen · 99 lines · 51 tokens per session scan A f915c37cc8ec

Subscribe to this mod's changes

core-tester is an agent published in the GitHub repository LukeRenton/explore-claude-code (309 stars, last pushed 21d ago), licensed MIT. It adds 51 tokens to every session and 1,027 once invoked, about $0.0003 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.

Related

Other agents, from other repositories