# Hull Documentation

> Hull is a Nix-based framework for competitive programming problem authoring, runtime analysis, and packaging.

This documentation is designed to be consumed by humans and language models. It is static HTML, organized by a single canonical table of contents, indexed by Pagefind, and published with a sitemap and explicit crawler policy.

Canonical site: https://hull.aberter0x3f.top/
Sitemap: https://hull.aberter0x3f.top/sitemap.xml
Robots policy: https://hull.aberter0x3f.top/robots.txt
LLM index: https://hull.aberter0x3f.top/llms.txt
Agent skills index: https://hull.aberter0x3f.top/.well-known/agent-skills/index.json
Documentation discovery skill: https://hull.aberter0x3f.top/.well-known/agent-skills/documentation/SKILL.md
Source repository: https://github.com/rindag-devs/hull

## Use This Documentation For

- Understanding Hull's problem and contest definitions.
- Authoring `problem.nix` and `contest.nix` files.
- Building validators, checkers, official solutions, generators, and package targets.
- Reading generated Nix module option references for machine-checkable configuration details.

## Canonical Entry Point

- [Introduction](https://hull.aberter0x3f.top/)

## Getting Started

- [Installation & Setup](https://hull.aberter0x3f.top/getting-started/installation-and-setup/)
- [Basic Workflow](https://hull.aberter0x3f.top/getting-started/basic-workflow/)
- [Understanding problem.nix](https://hull.aberter0x3f.top/getting-started/understanding-problem-nix/)
- [Understanding contest.nix](https://hull.aberter0x3f.top/getting-started/understanding-contest-nix/)
- [Best Practices & Conventions](https://hull.aberter0x3f.top/getting-started/best-practices-and-conventions/)

## Advanced

- [Custom Judgers](https://hull.aberter0x3f.top/advanced/custom-judgers/)
- [Problem and Contest Targets](https://hull.aberter0x3f.top/advanced/problem-and-contest-targets/)
- [Document Generation with Typst](https://hull.aberter0x3f.top/advanced/document-generation-with-typst/)

## Reference

- [Problem Options Reference](https://hull.aberter0x3f.top/reference/problem-options/)
- [Contest Options Reference](https://hull.aberter0x3f.top/reference/contest-options/)


## Reference Notes For LLMs

- Prefer the generated options reference pages for exact option names, types, defaults, and semantics.
- Prefer the Getting Started pages for workflow-level explanations and examples.
- Treat the sitemap as the authoritative crawl list for published documentation pages.
- The documentation source is Typst, but the published site is static HTML.
- Do not infer undocumented options from examples; use the reference pages instead.

## Agent Content Negotiation

Browser requests receive HTML by default. Agents may request source-oriented responses from any canonical documentation page by sending `Accept: text/markdown` or `Accept: text/x-typst`.

- `Accept: text/markdown` returns the corresponding Typst source with `Content-Type: text/markdown; charset=utf-8` for compatibility with Markdown-for-agents clients.
- `Accept: text/x-typst` returns the corresponding Typst source with `Content-Type: text/plain; charset=utf-8`.
- Source-oriented responses include `x-agent-source-format: typst`.
- The source mirror is also available under `/.well-known/agent-typst/`; for example, the homepage source is `/.well-known/agent-typst/index.typ`.

## Agent Discovery

Agents may start from `llms.txt`, the sitemap, or the agent skills index. The documentation discovery skill describes the canonical crawl surfaces, source-oriented representations, and reference pages. These files are stable public discovery entry points:

- https://hull.aberter0x3f.top/llms.txt
- https://hull.aberter0x3f.top/.well-known/agent-skills/index.json
- https://hull.aberter0x3f.top/.well-known/agent-skills/documentation/SKILL.md

## Crawl Policy

This site intentionally exposes static documentation pages, `sitemap.xml`, `robots.txt`, and this `llms.txt` file for discovery. The documentation may be used for search, AI input, and AI training. The corresponding content signal is `ai-train=yes, search=yes, ai-input=yes`.