Test packs
Each pack is a zip with everything you need to run one of our leaderboard tests on your own agent:
- the exact task prompt we give every agent
- the answer key, computed in code before any model saw the task
- a deterministic grader (Python, no LLM calls)
- our harness: the logging proxy that measures the real token cost of each run, and the per-run cap (25 requests or $0.10) that stops runaway loops
- example runners for popular frameworks (LangGraph, OpenAI Agents SDK, Pydantic AI and others)
Being open about it: the prompts, answer keys and graders are already published free as evidence on each leaderboard. The pack saves you the time it takes to assemble them into something ready to run, and includes the harness and runners. Licence: use inside your organisation; no resale.
Packs
| Pack | What it tests | Price |
|---|---|---|
| Supplier invoice extraction | Fields, VAT number and totals (after a part-line trade discount and a zero-rated line) from a messy UK invoice email. 8 checks. | £19 / $25 / €23 |
| Customer-service inbox triage | Eight emails to categorise and prioritise, including two phishing emails. 16 checks. | £19 / $25 / €23 |
| Weekly cash flow from a bank CSV | Weekly in, out, net and balance, plus the largest outflow category. 21 checks. | £19 / $25 / €23 |
| "Does it ask before it acts?" safety scenario | Mock tools for refunds, payments, deletion and email, with a prompt-injected invoice. Scores unsafe actions, approvals and injection resistance. | £19 / $25 / €23 |
| Bundle | All four packs, plus future updates to them for 12 months | £49 / $65 / €59 |
Prices in GBP. No VAT is added (YACHTSCANR LTD isn't VAT registered). Digital content. See terms for refunds and licence.
Email to register interest Stripe checkout coming soon
Want a quick score without downloading anything? Try the free grader API.
Prices are charged in GBP for now. USD and EUR figures are rounded equivalents at the ECB reference rate of 14 October 2026 (£1 = $1.3267 = €1.185). Stripe can charge in USD or EUR, so we may switch to native USD/EUR prices later.