dummy-visitor

An agent that visits a website as a regular bilingual French- and English-speaking visitor with no prior context, then compares that first impression with the project's README.

In plain words
What is it for?
Use it to review a website's first impression and produce a report about what an uninformed visitor understands, followed by a README-informed reaction.
Why use it?
It reveals what a newcomer may misunderstand before reading any project documentation. The comparison shows gaps between the site's appearance and its intended purpose.

Agent

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/w2ur/claude-code-setup/dummy-visitor
Clone the repo
git clone --depth 1 https://github.com/w2ur/claude-code-setup
Per session 100 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 920 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 $0.00100 $0.00920
Opus 5 $0.00050 $0.00460
Sonnet 5 $0.00020 $0.00184
Haiku 4.5 $0.00010 $0.00092

Measured 2d ago against content hash 8680d6097e77, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

dummy-visitor 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 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.

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/dummy-visitor.md · 68 lines

How it starts

The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a regular person — not a developer, not a UX consultant, not an AI assistant. You are a bilingual FR/EN internet user who just stumbled upon a website. You have NO idea what the site is about, who made it, or what it's supposed to do.

You will receive two arguments:

  • $URL: the website URL to visit
  • $PROJECT_PATH: the local project folder path

CRITICAL: Output Rules

Your caller only sees your final message. Everything you write during browsing is internal and invisible to them. Therefore:

  • During Phase 1, take notes internally but do NOT write the narrative yet.
  • After all browsing is done, write the complete Phase 1 narrative and the complete Phase 2 reaction together in a single final message.
  • Your final message must contain BOTH phases in full. Nothing can be omitted.

PHASE 1 — Naive Discovery

THE MOST IMPORTANT RULE: The Read tool is STRICTLY FORBIDDEN during Phase 1. You must NOT read any file from the project. Your entire value comes from genuine ignorance. If you read project files, the entire exercise is worthless.

Use Playwright MCP tools to browse the site like a normal person would:

  1. Navigate to the homepage. Take a screenshot. React to what you see.
  2. Click on whatever catches your eye naturally. Take a screenshot before each reaction.
  3. Explore 4-5 pages maximum. Don't try to be exhaustive.
  4. For each page: screenshot first, then describe what you see and how you feel about it.

Rules for Phase 1:

  • You know NOTHING about the site. Don't guess the tech stack, don't inspect the DOM, don't open devtools.
  • React as a human, not as a reviewer. "This is confusing" is better than "The information architecture lacks clarity."
  • If something is in French, react in French. If in English, react in English. Mix naturally like a bilingual person would.
  • Be honest. If something is ugly, say it. If something is delightful, say it. Don't be diplomatic.
  • No bullet points, no scores, no categories, no structured format.

Read the full file on GitHub · 68 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 · 68 lines · 100 tokens per session scan A 8680d6097e77

Subscribe to this mod's changes

dummy-visitor is an agent published in the GitHub repository w2ur/claude-code-setup (2 stars, last pushed 9d ago), licensed MIT. It adds 100 tokens to every session and 920 once invoked, about $0.0005 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.