Guide · Agent skill
How to use the Idea Ingest agent skill (idea-ingest)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The idea-ingest skill is an agent skill for gbrain, a personal or team knowledge base. When you share a link, article, tweet or idea, it tells the agent to fetch the content, file it as a page in the brain, write an analysis tied to what the brain already knows, and link related pages in both directions. This guide explains what the skill does, how to install it, how to trigger it, and what to check before using it. It assumes you have never used an agent skill before.
Idea Ingestgarrytan/gbrain · MIT Download the skill Quick start
- 1Download the ZIP. Get the idea-ingest ZIP from the skill's page in this directory.
- 2Unzip into a skills folder. Place the idea-ingest folder in ~/.claude/skills/ for personal use, or in a project's .claude/skills/ folder.
- 3Confirm the gbrain setup. Make sure your gbrain knowledge base has the search, query, get_page, put_page, add_link, add_timeline_entry and file_upload tools, and that the gbrain command-line tool is installed.
- 4Share a link with a clear request. Send a URL with a phrase such as "read this" or "save this", and say what you want the analysis to focus on.
- 5Review the new page. Check the filed page, its author credit and its cross-links, and fill in any gaps the agent reports.
- 6Sync the brain. Run gbrain sync so the index includes the new page.
What is an agent skill, and what does idea-ingest make the agent do?
An agent skill is a folder containing a SKILL.md file that tells an AI agent how to handle a specific kind of task. The agent reads the instructions when the task comes up. The idea-ingest skill covers one job: taking an item you share and putting it into your gbrain knowledge base.
For each item, the skill asks the agent to fetch the content, save it as a page, credit the author, and write an analysis that connects the item to existing projects, people and companies in the brain. It also keeps a copy of the raw source for reference.
- Fetches links, articles, tweets and ideas
- Saves a brain page with analysis, not just a summary
- Credits the author, either on an existing people page or inline on the ingested page
- Links related pages both ways
- Keeps the raw source for provenance
- If content cannot be fetched, saves a stub and asks you to paste the text
Install the idea-ingest skill
Download the idea-ingest ZIP from the skill's page in this directory and unzip it. The skill is a folder that contains SKILL.md. Put that folder in one of two places, depending on how you want to use it.
For a personal install across all your projects, place it in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/idea-ingest/
- Project install: .claude/skills/idea-ingest/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Check the requirements before you start
The skill depends on a working gbrain knowledge base. It uses the brain's search, query, get_page, put_page, add_link, add_timeline_entry and file_upload tools. Without those, it cannot file anything.
You also need the gbrain command-line tool for syncing and raw file upload, and access to a web fetch capability so the agent can retrieve links. The skill does not run scripts, but it does use the network.
- A gbrain knowledge base with the tools listed above
- The gbrain command-line tool
- Web fetch access to retrieve links
How do you trigger the skill with natural requests?
You do not need a special command. The skill is meant to activate when you share a link or URL, or say something like the phrases below. Be clear about what you want saved so the agent files it as intended.
Phrases that match the skill's triggers include "read this", "save this", "think about this" and "put this in brain".
- "Read this: https://example.com/article"
- "Save this and tell me how it relates to my current projects."
- "Think about this idea and put this in brain."
- Pasting a URL with "put this in brain"
What happens during an ingest?
The agent follows a fixed sequence. It fetches the content, uploads the raw source for provenance, checks the brain for an existing author page, files the item by its primary subject, writes the analysis, and then runs a sync so the index is updated.
Filing is by primary subject, not by format. A page about a person goes in people/, a company in companies/, a reusable framework in concepts/, and a raw data dump in sources/. The skill also checks for duplicate URLs and updates an existing page instead of creating a second one.
- Fetch the content (articles by web fetch, tweets by API, documents by PDF reader)
- Upload the raw source
- Look for an existing author page and update it, or credit the author inline
- Save by primary subject
- Write the analysis connecting to the brain
- Run gbrain sync
gbrain syncWhat to check for safety and privacy
The raw source is uploaded and stored in your knowledge base. If the brain is shared or synced to the cloud, anything you ingest may be stored there too. Do not use this skill for sensitive or private documents in that kind of store.
The skill uses the network to fetch links. It does not run scripts. If you only want a quick one-off summary and do not want pages created or edited in a persistent store, do not use this skill.
- Raw sources are uploaded and kept, so avoid private documents in shared or cloud-synced brains
- Network access is used to fetch links
- No scripts are run by this skill
Tips and troubleshooting
If the fetch fails because of a paywall, a 404 or a timeout, the agent saves a stub with the URL, metadata and reason, and asks you to paste the content. Pasting the text lets it complete the analysis.
If an upload is refused because the file is outside the allowed root or too large, the agent should tell you the limit and offer a link or excerpt instead. For video or podcast links, only metadata is ingested unless you supply a transcript.
- Paywall, 404 or timeout: paste the article text when asked
- Upload refused for size or path: use a link or an excerpt
- Video or podcast: provide a transcript for a full analysis
- Duplicate URL: the existing page is updated rather than duplicated
- Tweet threads are fetched as a single unit
What people use it for
- Filing articles you read so each one is connected to your active projects
- Collecting tweets or thread links and checking them against notes on people and companies
- Building a research notebook where each source is summarised against earlier material
- Keeping a team knowledge base current with reading that the team shares
Frequently asked questions
Is the idea-ingest skill free to use?
The skill is published under the MIT licence, as listed on this page. You still need a gbrain knowledge base and the gbrain command-line tool, and the source does not describe pricing for those.
What do I need to run idea-ingest?
You need a gbrain knowledge base with the search, query, get_page, put_page, add_link, add_timeline_entry and file_upload tools, the gbrain command-line tool, and web fetch access for links. The skill does not run scripts.
Where do I install an agent skill like idea-ingest?
Unzip the folder into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
What happens if the agent cannot fetch a link?
It saves a stub page with the URL, metadata and the reason for the failure, and asks you to paste the content. Once you paste it, the agent can complete the analysis.
Does idea-ingest create a page for every author?
No. It creates a new people page only when the author passes the notability gate and there are at least two independent facts about them beyond the single item. Otherwise it credits the author inline on the ingested page.
Is it safe to save private documents with this skill?
Be cautious. The raw source is uploaded and stored in the knowledge base, so private or sensitive documents should not go into a shared or cloud-synced brain.
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
Use idea-ingest if you already run a gbrain knowledge base and want links you collect filed, credited and connected to your existing notes without manual work. Pick something else if you do not use gbrain, need a quick summary without creating stored pages, or plan to ingest sensitive material into a shared or cloud-synced store.
Download the skill