Skip to content
Get packs

Guide · Agent skill

How to use the Brain Taxonomist agent skill (brain-taxonomist)

Updated · written with AI from the project's own documentation and checked against our catalogue data

Brain Taxonomist is an agent skill for people who keep a personal or team knowledge base called a gbrain. Before the agent creates a new page, this skill checks the brain's active schema, which is the set of rules describing page types and folders, and recommends where the page should go. After this guide you will know how to install the skill, ask for filing checks in plain language, and run the drift audits that look for pages in the wrong or undeclared types.

Brain Taxonomistgarrytan/gbrain · MIT

Quick start

  1. 1Install gbrain and jq. Make sure the gbrain command-line tool with schema subcommands and jq are available on your machine.
  2. 2Confirm an active schema pack. Run gbrain schema show --json to check that your knowledge base has an active schema pack.
  3. 3Place the skill folder. Unzip the downloaded ZIP into ~/.claude/skills/ or a project's .claude/skills/ folder so SKILL.md sits inside brain-taxonomist.
  4. 4Ask a filing question. Describe the new page in plain language and ask where it should go, for example with the phrase where does this brain page go.
  5. 5Review the recommendation. Check the file path, matched page type, and active pack in the output before the agent writes anything.
  6. 6Run a drift audit. Ask for a taxonomy check, or run gbrain schema review-orphans --json and gbrain schema lint --with-db --json to find undeclared types.

What an agent skill is and what this one does

An agent skill is a folder containing a SKILL.md file that tells an AI agent when and how to perform a task. This skill, brain-taxonomist, acts as a filing check for a gbrain. It does not have its own directory list. Every decision is read from the active schema pack by running the gbrain command-line tool.

It works in two ways. The gate function is consulted before a new brain page is created, so the page lands in the correct path from the start. The drift function is a periodic scan that finds pages that have outgrown their current location or have drifted into a type the schema does not declare.

  • Recommends a file path for new pages based on the primary subject (person, organization, event, concept, media, or bulk source data).
  • Reads the active schema pack, so a custom pack gets recommendations based on your own taxonomy.
  • Flags ambiguous cases for you to decide instead of silently picking a folder.
  • Reports undeclared and orphaned page types during audits.

Install the brain-taxonomist skill

Download the ZIP for this skill from this page and unzip it. The skill is delivered as a folder that contains SKILL.md. Put the unzipped folder in one of these locations, depending on whether you want it for all projects or one project:

You can also install it in the Claude apps by uploading the skill under Settings → Capabilities → Skills.

  • Personal (all projects): ~/.claude/skills/brain-taxonomist/
  • Single project: .claude/skills/brain-taxonomist/ inside your project folder
  • Claude apps: Settings → Capabilities → Skills, then upload the skill
~/.claude/skills/brain-taxonomist/SKILL.md

Check the requirements before you start

The skill depends entirely on the gbrain command-line tool and an existing knowledge base. It does nothing useful on its own. The skill also needs jq to filter the JSON output of gbrain schema show. Confirm that gbrain supports the schema subcommands the skill uses: schema show, review-orphans, and lint.

  • gbrain command-line tool with the schema subcommands (schema show, review-orphans, lint)
  • jq, used to filter the JSON output of gbrain schema show
  • An existing gbrain knowledge base with an active schema pack
  • Scripts: the skill runs scripts. Network: the skill does not use the network.
gbrain schema show --json | jq '.page_types[] | select(.primitive == "entity")'

Trigger the skill with natural requests

You do not need special syntax. Ask in ordinary language and the agent will load the skill when your request matches one of its trigger phrases. The skill is meant to run before you create a new page, before a bulk import, or when the subject of a note is unclear.

Trigger phrases from the skill include the examples below. You can also name the skill directly with brain taxonomist.

  • "Where does this brain page go?"
  • "File this in the brain."
  • "Taxonomy check."
  • "Refile brain page."
  • "Which directory does this page go?"

Use it for filing and drift audits

For a new page, describe the subject and ask where it belongs. The skill walks through the decision steps in order: is it a named person, a named organization, a time-bounded event, a reusable mental model, raw media, or bulk source data. It then looks up the matching directory in your active pack and returns a recommendation with a short reasoning trail.

For audits, ask for a taxonomy check. The skill runs two commands that report untyped pages and pages whose type the pack neither declares nor aliases. If more than 10% of a source is untyped, the skill suggests running schema detect and then review-candidates.

  • Example request: "Where does this brain page go? It is a note about a meeting with a supplier."
  • Example request: "Run a taxonomy check on the brain."
  • Example request: "Refile brain page for the Q3 reading notes; the subject is a book."
gbrain schema review-orphans --json gbrain schema lint --with-db --json

Read the output and decide

The skill returns an advisory block. It names the file path, the matched page type, the active pack, and the source. When the pack has no matching type, the skill does not guess. It tells you that no match exists and suggests the next commands to run.

The skill never changes the active pack. If a schema change is needed, it hands the decision to you so you can agree before anything is modified.

  • File at: the recommended path and slug
  • Matched page_type and primitive
  • Active pack name and version
  • Source ID, relevant when you use per-source overrides

Safety notes and tips

The skill is advisory. It recommends paths but the agent still writes files, so unchecked use can misfile content. Review the recommendation before you accept it, and do not use the skill as a reason to hand-edit or delete pages without checking them.

Ask for a filing check before you write new pages, not after. Keep the same content and you should get the same path, which makes results repeatable. If a page type is missing from your pack, declare it or rewrite the page with an existing type rather than accepting the closest fit.

  • Review every recommended path before the agent writes the file.
  • Use --source <id> when you run a multi-brain setup with different packs per source.
  • Run the drift audit periodically to catch undeclared types early.
  • Do not hand-edit the schema from this skill; route schema changes to EIIRP or schema-author.

What people use it for

  • Filing a new meeting note under the correct temporal directory instead of a general folder.
  • Deciding whether a new article belongs under media or concept in a custom schema pack.
  • Keeping separate filing rules for two brains that use different schema packs per source.
  • Auditing a brain before a bulk import to find pages with types the schema does not declare.
  • Filing book summaries into the category folder defined by the active pack.

Frequently asked questions

Is brain-taxonomist free to use and what is its licence?

The item is listed under the MIT licence. The skill itself is advisory and runs locally, but it requires the gbrain command-line tool and an existing knowledge base to work at all.

How do I install the brain-taxonomist skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.

What do I need to run brain-taxonomist?

You need the gbrain command-line tool with the schema subcommands schema show, review-orphans, and lint, plus jq to filter the JSON output. You also need an existing gbrain knowledge base with an active schema pack.

Does brain-taxonomist work without gbrain?

No. The skill depends entirely on the gbrain command-line tool and its schema. Without gbrain it does nothing useful.

Does the skill change or delete my pages?

No, it is advisory and only recommends paths. The agent still writes files, so review each recommendation before accepting it. The skill is designed never to change the active pack itself.

How does brain-taxonomist compare to other ways of organising notes?

Its main difference is that it reads your active schema pack rather than using a fixed directory table, so recommendations follow your own taxonomy. If you do not use gbrain, this skill is not a fit and you should look for a note-filing tool built for your own system.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Brain Taxonomist is a good choice if you already run a gbrain and want new pages filed consistently, especially if you keep several schema packs or need to audit undeclared page types. It is not useful if you do not use gbrain, and you should not rely on it to clean up pages without reviewing its recommendations.

Download the skill