Skip to content

Guide · Agent skill

How to use the Retrieval Reflex agent skill (retrieval-reflex)

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

Retrieval Reflex is an agent skill, a set of instructions that teaches an AI agent a specific behaviour. This one makes the agent open the stored page for a named person, company, project or deal in your GBrain knowledge base before it answers from memory. This guide explains what the skill does, how to install it, how to trigger it, and what it needs to work well.

Retrieval Reflexgarrytan/gbrain · MIT

Quick start

  1. 1Get a GBrain instance. Set up GBrain with brain pages and make sure get_page, query, traverse_graph and get_backlinks are available through MCP or the gbrain CLI.
  2. 2Download the ZIP. Download the retrieval-reflex ZIP from the item page.
  3. 3Install the skill. Unzip it into ~/.claude/skills/ for all projects or into a project's .claude/skills/ folder, or upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
  4. 4Discuss a named entity. Talk about a person, company, project or deal as you normally would, with no special command needed.
  5. 5Check the reply. Confirm the agent opened the stored page before giving specifics; if it did not, ask it to check.

What the Retrieval Reflex skill makes the agent do

The skill copies a habit people already have: before speaking about someone, you check your notes. When a named entity such as a person, company, project, place or deal becomes the subject of a conversation, the agent looks up its page in your brain (your personal knowledge store) instead of relying on memory.

The skill also limits how much the agent reads. It escalates only as far as the task needs: first a short pointer with the slug and a one-line summary, then the full page when details matter, and linked neighbours only when relationship context is required. It does not load everyone in your contacts at once.

  • Pointer only: enough when the task just needs identity.
  • Full page: opened when the entity is the subject or specifics matter.
  • Linked neighbours: pulled only when relationships are part of the answer.

Install the retrieval-reflex skill

Download the ZIP for the retrieval-reflex skill from the item page. Then choose one of two ways to install it.

For Claude Code, unzip the folder into ~/.claude/skills/ to make it available everywhere, or into a project's .claude/skills/ folder to make it available only in that project. The skill is a folder containing SKILL.md.

In the Claude apps, you can upload the ZIP instead under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/
  • Project install: .claude/skills/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

What the skill needs before it works

The skill needs a GBrain instance with brain pages and four operations available: get_page, query, traverse_graph and get_backlinks. You can reach these through MCP or through the gbrain CLI. Over the CLI, traverse_graph and get_backlinks are exposed as gbrain graph and gbrain backlinks.

Without a populated brain, lookups will return little or nothing, so the skill will not add much value yet.

  • A running GBrain instance with brain pages
  • Operations: get_page, query, traverse_graph, get_backlinks (via MCP or the gbrain CLI)
  • Note: the source lists no scripts and no network use for this skill
get_page <slug>

How to trigger the skill with natural requests

You do not need a special command. The skill fires on the ambient reflex of retrieving before answering. It is triggered when a named entity becomes the subject of the exchange, when a brain-page pointer appears in context, or when the agent is about to state a detail about an entity from memory.

Direct lookup questions such as 'who is' or 'tell me about' are routed to the separate query skill instead. If you want an explicit search, use that skill.

  • Works on: 'Before I reply to Maria about the renewal, what do we have on her?'
  • Works on: a request to draft a note about a company you have discussed for several messages
  • Use the query skill for: direct lookups like 'who is Daniel Chen?'

Safety notes: automatic reads of personal records

The skill reads stored pages without asking you each time. That is the point, but it means it is not suited to contexts where automatic reads of personal records are unwanted.

The skill is marked as non-mutating and does not write pages, according to its metadata. It runs no scripts and uses no network, as listed in the item details.

  • It reads brain pages automatically when triggers match.
  • It does not write or change pages (writes_pages is false).
  • Skip it in sensitive settings where you do not want records read without a prompt.

Tips for getting good results

Keep the brain populated with the people, companies and deals you discuss most, because the skill can only retrieve what exists. Ask about one entity at a time when you can, since the skill is designed to resolve only the names a task needs.

If the agent states a detail without checking, say 'let me check' or ask it to open the page. The skill's own guidance is to verify before asserting specifics.

  • Populate the brain first; empty pages return nothing useful.
  • Name the entity clearly so it is recognised as the subject.
  • Ask for the page explicitly if a conversation has gone on without it being opened.

What people use it for

  • Before drafting a follow-up to a contact, the agent opens their page to recall the last discussion.
  • When a deal comes up mid-conversation, the agent checks its status in the brain before stating it.
  • When a project is mentioned by name, the agent pulls the page to confirm history before answering.
  • When relationships matter, the agent pulls linked neighbours to see who is connected to whom.

Frequently asked questions

Is the retrieval-reflex skill free to use?

The item is licensed under MIT, as listed on the item page. The source does not state any price, so check the item page for current availability.

What do I need to run the retrieval-reflex skill?

You need a GBrain instance with brain pages and the get_page, query, traverse_graph and get_backlinks operations, available through MCP or the gbrain CLI. The skill runs no scripts and uses no network.

Where do I install the retrieval-reflex skill?

Unzip the ZIP into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

Why does the agent not check my brain when I ask 'who is' someone?

Direct lookup questions such as 'who is' or 'tell me about' are routed to the separate query skill. This skill fires on the ambient habit of retrieving before answering, so use the query skill for explicit searches.

Why does the skill return nothing?

The skill can only retrieve pages that exist. If your brain has no pages for the entity, or the GBrain operations are not connected, lookups will return little or nothing.

How does this compare to other memory skills?

The source does not compare it with other tools. It is aimed at automatic retrieval before answering, and the source points to the query and brain-ops skills for search and brain operations.

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

Choose this skill if you keep a GBrain knowledge base and want your agent to check it automatically before speaking about people, companies or deals. Pick something else if you only want explicit searches, have no populated brain, or do not want an agent reading personal records without asking each time.

Download the skill