> For the complete documentation index, see [llms.txt](https://nfa.gitbook.io/nfa-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://nfa.gitbook.io/nfa-docs/the-forecasting-engine/cruxes-and-verification.md).

# Cruxes and verification

## Cruxes and verification

Most forecasts turn on one fact. NFA makes the engine find that fact, argue about it, and check it, rather than letting a decisive but unverified claim slip through.

### The crux

After the opening statements, the judge names the **single most decisive unresolved question** in the debate. That question is the crux. The engine then runs one targeted search on it, and the evidence it finds is fed into the rebuttal rounds, so the advocates argue the point that actually matters instead of talking past each other.

The crux question and a short, price-redacted summary of its evidence are surfaced in the run's trace, and sealed so the run can be replayed.

### Ruling-time verification

There is a specific failure mode worth guarding against: an advocate makes a claim that would change the verdict, it is true, but it appears in neither the shared corpus nor any advocate's dossier. A judge with no way to check it either ignores a decisive truth or accepts a decisive fiction.

So just before ruling, the engine runs a verification pass:

{% stepper %}
{% step %}

#### Name the claim

The judge names at most one advocate claim that could change the verdict and is supported by neither the corpus nor any dossier.
{% endstep %}

{% step %}

#### Check it

One targeted search tries to confirm or refute that claim.
{% endstep %}

{% step %}

#### Fold the result in

If evidence is found, it is appended to the shared brief as verification evidence and the judge rules with it in hand. If the search comes back empty, the claim is recorded as an unverified crux.
{% endstep %}
{% endstepper %}

{% hint style="success" %}
Every run seals a verification record: the claim that was checked, whether it was searched, and whether evidence was found. Unverified decisive claims are listed explicitly, so a forecast that rested on something the engine could not confirm is visible rather than hidden.
{% endhint %}

This pass exists because of real misses where a true, decisive, late-breaking fact was absent from the corpus. Verifying the one claim that would flip the ruling is a cheap, targeted way to catch exactly that class of error.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://nfa.gitbook.io/nfa-docs/the-forecasting-engine/cruxes-and-verification.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.