Tutorial

Hyper-Extract hands-on guide

Chapter 1 free: install skills and start with brainstorm. Later: template types, install he CLI for first extract, failures (paid).

Open the free related Skill package

Chapter 1 · Install the skills, start with brainstorm

Plain talk: Hyper-Extract is a document grinder—messy text in, structured knowledge out. You still need a mold (YAML template). This free package teaches the assistant to design molds; pressing the switch needs the CLI separately.

This chapter: install skills → copy-paste prompt → get a type recommendation / design draft.

Package free (Apache-2.0). Later chapters cover installing he and the first parse.


Where people get stuck

  1. Unzipping the whole GitHub repo as a Skill with the wrong folder name;
  2. Asking to “extract the whole book” without naming fields;
  3. Expecting offline extraction from the skills zip alone—no CLI, no API key.

1. Install (about 5 minutes)

  1. SkillBazaar → free Skill “Hyper-Extract Template Designer” → download zip.
  2. Top folder must be hyper-extract under your skills path, e.g. ~/.cursor/skills/hyper-extract, with root SKILL.md plus brainstorm/, record-designer/, graph-designer/, etc.
  3. New chat. @hyper-extract or “Use Hyper-Extract skills to design an extraction template.”

Check: ask what brainstorm vs graph-designer do. Clear “choose type, then design graph template” = installed.


2. Copy-paste prompt

Use Hyper-Extract brainstorm: I want person–work relations from a biography Markdown.
Confirm inputs/outputs, recommend model / list / set / graph / hypergraph, and give a short design draft.
Do not emit final YAML yet, and do not pretend he parse already ran.

If it misbehaves

  • Dumps a huge YAML → “Stay in brainstorm; no full template before type confirm.”
  • Fakes parse results → “Design only; no CLI run required in this chapter.”
  • Ignores child skills → “Follow hyper-extract: brainstorm → designer.”

3. Checklist

  • hyper-extract/SKILL.md and at least one child skill on disk
  • Assistant gave a type suggestion and short draft
  • You know skills ≠ installed he
  • Next time you’ll state input shape and required output fields

Scroll down for record vs graph types, CLI first run, and failures.

Unlock later chapters

Chapter 1 is fully free. Pay to continue with layout rules and advanced steps. (¥1.90)

  • Coming up
  • Chapter 2 · Which template: records or graphs?
  • Chapter 3 · Install the CLI, first extract, failures