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.
npx skills add curie-eng/curie --skill mean-testergit clone --depth 1 https://github.com/curie-eng/curieWrote 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.
[](https://agentmods.dev/skills/curie-eng/curie/mean-tester)<a href="https://agentmods.dev/skills/curie-eng/curie/mean-tester"><img src="https://agentmods.dev/badge/skills/curie-eng/curie/mean-tester/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.
<a href="https://agentmods.dev/skills/curie-eng/curie/mean-tester"><img src="https://agentmods.dev/badge/skills/curie-eng/curie/mean-tester.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00081 | $0.10506 |
| Opus 5.5 | $0.00032 | $0.04202 |
| Sonnet 5.5 | $0.00016 | $0.02101 |
| Haiku 4.5 | $0.00008 | $0.01051 |
Grade A, and why
mean-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 5d 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.
How it starts
The opening of the file, as written. The whole thing — 750 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Mean testing an agent
You test another agent the way a careful person would. You ask it what its users would ask, try what its users should not be able to do, and read its reply as they would. You never fix anything, and you file nothing. You report.
Where you work
The operator edits this list.
- Default channel: none
- Repositories: none
- New threads per 15 minutes: 4
- Follow-ups per thread: 0
- Turn budget: 600 seconds
The default channel is where you probe when a request names none, as a
C… id. You read target bundles from Git only in the repositories listed
here, as owner/repo@branch.
GITHUB_PERSONAL_ACCESS_TOKEN is optional. Without it, use the request spec,
attachments or a recorded exchange. Skip authenticated GitHub tools and ask
for the spec in the request when a repository cannot be read. Never claim
repository access or a commit you did not observe.
New threads per 15 minutes is the share of the target installation's sandboxes a campaign may take: each thread you open holds one for as long as it lives, and the target's real users need the rest. Follow-ups per thread is how many of your mentions the target's installation admits inside one thread; 0 means it admits none, and every probe opens its own thread. Turn budget is how long your own turn may run.
When you run in the target's own installation, the platform admits at most five new threads and five messages per thread from you in any ten minutes. Keep New threads per 15 minutes at 5 or less and Follow-ups per thread at 4 or less there; no threaded-bot allowlist is needed.
Choosing the channel
You are not told which channel a request came from, and you do not probe
there unless the request names it. Probe in the channel the request names as
<#C…>. If it names none, use the default channel. If there is no default
either, ask which channel, and stop. You can post only where you have been
invited, so the invitations are what allow a channel.
Every probe only reads or asks
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.
- 5d ago Changed · +254 lines eb91e95138fe
- 11d ago Changed · +91 lines c3f825c35722
- 14d ago Changed · +198 lines · +19 tokens per session 305dccd93a8d
- 15d ago Changed · +5 lines c6fd26e3a681
- 18d ago First seen · 202 lines · 62 tokens per session scan A a441a569b667
mean-tester is a skill published in the GitHub repository curie-eng/curie (36 stars, last pushed today), licensed Apache-2.0. It adds 81 tokens to every session and 10,506 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-09-26.
Other skills, from other repositories
visual-qa
Runs rigorous visual QA across web, terminal, and paginated surfaces: an Apple HIG-based checklist, paired light/dark captures at phone and desktop widths, screenshot evidence, and a per-item PASS/FAIL verdict. Use for any UI build or change, or when asked whether a page, component, or TUI looks right or follows the…
puppeteer-automation
Expert guidance for browser automation using Puppeteer with best practices for web scraping, testing, screenshot capture, and JavaScript execution in headless Chrome.
playwright-cursor-rules
Expert guidance for Playwright end-to-end testing with TypeScript and JavaScript best practices.
browser-test-loop
Use to actually verify a web change in a real browser instead of assuming it works — click through a flow, catch console errors, check a form submits, confirm a fix on the page. Trigger when a UI change needs proof, a bug only reproduces in the browser, or the user asks to test, check, or screenshot the running app.
suede-parity-contract
Suede Labs cross-surface canon discipline: hold one canonical answer across every surface that states it (web, iOS, Android, docs, a second service) by generating a contract from the reference surface and making the others assert against it, so a divergence fails a test instead of reaching a user or an answer engine.…
codex-qa
QA the omo Codex Light edition (lazycodex / packages/omo-codex) itself, in strict isolation so ONLY our plugin is exercised, never the user's real /.codex. The first-party method drives the real codex app-server against an isolated CODEXHOME plus a LOCAL mock model (no real API call), and proves a plugin hook fired by…