Order a Task Risk Assessment

The form below is the entire ordering process: six fields, one submission, one deterministic result. Pricing runs from EUR 1.90 across three fixed tiers — see the full breakdown at /pricing.

There is no second page. There is no account to create, no cart to review, no confirmation email to wait for. Six fields go in and a numbered certificate comes out, and the whole of the distance between those two events is one button.

A single illuminated glass intake slot set into a white console, one printed card suspended above it.
  1. Choose a tier
    Three fixed tiers, priced in advance, printed as text.
  2. Describe the task
    Six fields, one submission, no account, no call.
  3. Receive a certificate
    Numbered, permanent, verifiable, issued immediately.

Checkout status

Checkout activation is in progress. Until it completes, this form validates the parameters submitted and reports onboarding status instead of taking payment. No charge is made and no certificate is issued in that state. The form below can still be completed, and the parameters it returns are the ones a live order would use.

The model identifier of the agent under assessment — for example claude-opus-5 or gpt-5.6-sol. The certificate that results describes this configuration, so the identifier entered here is what the assessment is about. Registered identifiers and their manufacturers are listed in the table below.

Format YYYY-MM, e.g. 2026-05 — the training cutoff of the agent named in the Model identifier field above. Month precision is sufficient; a day is not requested and would not be used. Registered configurations and the cutoffs on file for them are listed in the table below.

Sampling temperature used for this configuration, to three decimals exactly — the value is hashed as a string, so "0.7" is not accepted where "0.700" is. 0.700 is the Registry's common default and is preselected; use it when the configuration under assessment does not specify a temperature of its own.

The category that best describes the task being assessed. This is a fixed list rather than free text, because no language model is permitted anywhere on the delivery path.

Between 10 and 500 characters. The task description is used to compute a one-way digest and is immediately discarded. It is never stored, never logged, never published, and never repeated back — including in an error message on this page.

One per line, up to 10. Use generic tool names, not internal system names. Listed tools receive individual compatibility ratings on the Extended and Full Chart tiers; leaving this field empty omits that section and changes nothing else.

Assessment tier

What is sold here is an automated task risk assessment: a fixed, published method applied to the parameters entered above, with no human review and no language model on the delivery path. It is not advice, and it is not a guarantee of any task outcome.

Registered configurations

The three fields above ask for a specific agent's configuration. The table below lists the configurations the Registry already has on file, for anyone completing the form who needs to look one up rather than supply it from memory.

The Registry publishes reference values for ten configurations in advance, each model identifier paired with its registry default training cutoff. Find the agent's manufacturer below and read across the row for the cutoff on file.

Model Manufacturer Registry default training cutoff
Claude Fable 5 Anthropic 2026-01
Claude Opus 5 Anthropic 2026-05
Claude Sonnet 5 Anthropic 2026-01
GPT-5.6 Sol OpenAI 2026-02
GPT-5.6 Terra OpenAI 2026-02
Grok 4.6 xAI 2026-02
Gemini 3.7 Flash Google DeepMind 2025-01
Kimi K3 Moonshot AI 2026-03
DeepSeek V4 Pro DeepSeek 2025-12
Qwen3.8-Max Alibaba Cloud 2026-03

Where a manufacturer has not published a training cutoff, the value shown above is the Registry's own assumption, stated for consistency rather than as a manufacturer claim.

This table is not exhaustive and does not limit what the form above will accept: any model identifier and training cutoff may be submitted, whether or not it appears here. These ten are simply the configurations for which the Registry publishes a reference cutoff ahead of time.

What happens after submission

Submission triggers three things in order: validation, computation, issuance. Validation canonicalises the six fields and rejects anything outside the documented constraints. Computation runs AFR-1 over the canonical values, with no language model and no network call involved. Issuance writes a numbered certificate to a permanent address and returns it. Three stages, one request, no queue between them.

On the task description It's worth noting that the task description does not survive step one. It is reduced to a digest, the digest is used, and the original text is gone before computation begins.

Agents

Structured pricing for this form's tiers is available at /pricing.json. A full integration guide — request shapes, field constraints, and purchase steps — is at /agents.

One task, one price, one certificate

Standard Assessment EUR 1.90 · Extended Assessment EUR 4.90 · Full Chart Assessment EUR 14.90. Every tier carries the verdict and the recommended execution window. And that is not all — the Extended Assessment additionally carries the 24-hour outlook, three named risk factors and tool compatibility ratings, while the Full Chart Assessment carries the permanent chart on top of those. One charge, one document, no renewal. The full breakdown is at /pricing.