How to use the Enrich Brain Pages agent skill

What is an agent skill and what does enrich do?

An agent skill is a folder with a SKILL.md file that gives an AI agent instructions for one kind of task. When a request matches the skill's triggers, the agent follows those instructions instead of improvising.

The enrich skill covers entity enrichment. When a person or company is mentioned in your work, the agent checks whether your brain already has a page for them, looks up new information, and writes or updates a page. Each page gets a State section of hard facts with inline citations, a dated timeline, and links back to the pages that mention them.

• Scales effort by importance using three tiers: Tier 1 full, Tier 2 medium, Tier 3 minimal.
• Checks the brain first with search and query before going to the web.
• Saves raw API responses so the compiled notes can be traced back to their source.
• Skips random mentions, bot or spam accounts, and entities with no link to your work.
• Does not re-enrich a page enriched within the past week unless new signal justifies it.

Install the enrich skill in Claude

Download the enrich skill as a ZIP from the skill's page in this directory. Then choose one of two install routes.

For a personal install, unzip the file into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder. Check that the enrich folder contains SKILL.md directly inside it.

You can also upload the ZIP in the Claude apps. Open Settings, go to Capabilities, then Skills, and add the file there.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside the project folder
• Claude apps: Settings → Capabilities → Skills

What you need before it works

The skill depends on a gbrain brain that exposes page tools. The source lists get_page, put_page, search, query, add_link, add_timeline_entry and get_backlinks. Without these the skill does nothing useful.

Other services are optional. A web research tool such as Perplexity, Brave Search or Exa enables deeper research in Tiers 1 and 2. Enrichment APIs such as Crustdata, Proxycurl, People Data Labs, Crunchbase or PitchBook are used for Tier 1, and each needs its own API key. Social media access helps when you know someone's handle.

• Required: a gbrain brain with the page tools listed above
• Optional: a web research tool for deeper tiers
• Optional: enrichment API accounts, each with its own key, for Tier 1
• Optional: social media access where handles are known

How to trigger the skill with natural requests

You do not need special syntax. The skill's triggers include the phrases enrich, create person page, update company page, who is this person and look up this company. Ask in plain sentences that name the entity and what you want.

The skill also runs as part of ingest work. The source says an entity appearing in a meeting transcript or email, or a new contact with significant context, is a reason to enrich.

• Look up this company: Acme Robotics, they just hired a new CTO.
• Who is this person? Priya Natarajan from the investor call.
• Create person page for Tom Liu, we met at the conference and he works on developer tools.
• Update company page for Northwind with the funding news from this week.

What the agent does step by step

The protocol has seven steps. The agent extracts entities from your input, searches the brain for each one, and then either updates an existing page or creates a new one. It pulls signal from the source, runs external lookups in priority order, saves raw data, and writes to the brain.

For a new page, the agent first resolves the entity's type and directory from the brain's active schema pack rather than assuming person or company. If no declared type fits, it says so and asks. Empty sections are marked [No data yet] instead of filled with boilerplate. When the agent meets a contradiction between new and existing information, it flags both citations.

• Briefly: identify, check brain, extract signal, look up sources, save raw data, write.
• Timeline entries are reverse-chronological and append-only.
• Compiled truth changes only when new signal materially changes the picture.
• User-written assessments are not overwritten by API boilerplate.
• Every mention of a page-backed entity creates a back-link from that entity's page.

Safety notes: scripts and network use

The skill's listing says it does not run scripts and does use the network. Its web research, social lookup and enrichment steps send requests to outside services, and the Tier 1 APIs require your own API keys.

The skill does not address privacy or consent for looking up private individuals beyond skipping spam and unrelated mentions. Use it for people with a clear business relationship to you. Web and API results can be wrong, and the skill relies on the agent's judgement about sources, so check anything important before you act on it.

• No scripts are run by the skill.
• Network calls go to web research, social and enrichment services you configure.
• Keep API keys for each service yourself.
• Verify facts that matter, since sources may be wrong or outdated.