Guide · Agent skill
How to use the Company brain follow-up agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The company-brain-follow-up-agent skill tells an AI agent how to run the follow-up agent cookbook from the cognee repository. It takes your latest Granola call, works out the next steps with an owner, team, due date and the existing Linear issue for each, then prints them or posts them to Slack. This guide explains what the skill does, how to install it, how to trigger it and what it needs before you run it.
Company brain follow-up agenttopoteretes/cognee · Apache-2.0 Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it so the folder contains SKILL.md.
- 2Place the skill. Move the folder to ~/.claude/skills/ for personal use, or to a project's .claude/skills/ folder. You can also upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Get the cookbook. Clone the cognee repository so examples/cookbooks/company_brain/follow_up_agent/ is available, and install uv and Python.
- 4Set LLM_API_KEY. Provide an LLM_API_KEY for the language model. This is the only required credential for a sample run.
- 5Run the check. Run the --check command from the repository root and fix any line marked MISSING before continuing.
- 6Run the agent. Run follow_up_agent.py and read the steps, the call they come from and whether they were posted.
What the skill makes the agent do
Agent skills are folders of instructions that an AI agent loads when a task matches them. This skill covers one task: following up on a meeting. The agent runs the follow-up agent cookbook in the cognee repository. The cookbook stores your Granola calls, Linear issues and Gmail inbox in a cognee dataset called follow_up_agent, then works out the next steps of your latest call.
For each next step, the output includes the owner, the team, a due date and the Linear issue that already tracks it. The steps are then either printed or posted to Slack, depending on whether Slack is set up. Gmail is read only.
Install the skill from the ZIP
Download the ZIP for this skill from this page and unzip it. The folder inside is named company-brain-follow-up-agent and contains a SKILL.md file.
For a personal install, put the folder in ~/.claude/skills/. To limit it to one project, put it in that project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/company-brain-follow-up-agent/
- Project only: .claude/skills/company-brain-follow-up-agent/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Set up the cookbook before a real run
The skill runs commands from the repository root, so you need a clone of the cognee repository with the cookbook at examples/cookbooks/company_brain/follow_up_agent/. The cookbook's README explains setup and lists what each script does. You need uv and Python, plus an LLM_API_KEY for the language model.
Granola, Linear and Gmail are optional. Without them, the skill runs on built-in sample data for a fictional call. Gmail needs credentials.json and token.json. Slack is needed only if you want messages posted.
- uv and Python (the cookbook runs through uv run)
- LLM_API_KEY for the language model (required)
- Granola access for real call data (optional)
- Linear access for issue data (optional)
- Gmail credentials.json and token.json, read-only (optional)
- Slack setup to post messages (optional)
Check your setup first
Run the check before anything else. It lists what is missing and what will be skipped. Exit code 0 means you can continue. Exit code 2 means a required item is missing; each line starting with [setup] MISSING: names one fix, and the cookbook README's "What it needs" table explains it.
uv run python examples/cookbooks/company_brain/follow_up_agent/follow_up_agent.py --checkRun the follow-up agent
Once the check passes, run the agent. The output says which call the steps come from, which sources were ingested and whether they were posted to Slack. Add --days N to set a time range, or --no-linear and --no-email to leave out a source.
Without your own accounts, the run uses sample data and never posts to Slack. Add --sample to use the sample even when your accounts are set up.
- --days N: time range for calls
- --no-linear: skip Linear issues
- --no-email: skip Gmail
- --sample: use sample data even when accounts are set up
uv run python examples/cookbooks/company_brain/follow_up_agent/follow_up_agent.pyTrigger the skill with natural requests
The skill loads when a request matches its description, which is about next steps or action items from the latest call. You do not need to name the skill. Requests like these should match:
- What are the next steps from my latest call?
- Turn my last Granola call into action items.
- Who owns the follow-ups from today's meeting, and which Linear issues track them?
- Post the action items from my latest call to Slack.
Safety notes: scripts and network
This skill runs scripts and uses the network. Every real run reads your calls, issues and mailbox and sends that data through an LLM, which uses credits. Gmail access is read only. Posts go only to Slack, and only when Slack is set up. Do not share the contents of credentials.json, token.json or .env.
A sample run clears the follow_up_agent dataset before ingesting. On your own data, the --clear flag forgets everything the dataset holds, so use it only when you want to start over. Avoid --ui unless you want to browse the graph, because it keeps running until you press Ctrl+C.
- Gmail is read only
- Slack posts only when Slack is set up
- Real runs use LLM credits and read your data
- --clear forgets the whole follow_up_agent dataset
Clean up the stored data
To remove what the cookbook stored in its dataset, run the command below from the repository root.
uv run cognee-cli forget --dataset follow_up_agentWhat people use it for
- Turning the action items from your latest Granola call into a list with owners and due dates.
- Finding the existing Linear issue behind each follow-up so you do not create duplicates.
- Posting the agreed next steps to a Slack channel after a team call, when Slack is set up.
- Trying the full flow on built-in sample data before connecting real accounts.
- Studying the cognee cookbook as an example of a memory-backed agent workflow.
Frequently asked questions
How do I install the company brain follow-up agent skill?
Download the ZIP from this page and unzip it. Put the company-brain-follow-up-agent folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder, or upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
Is the skill free to use?
The skill is published under the Apache-2.0 licence. Real runs use LLM credits through your LLM_API_KEY, so the cost depends on your language model provider. The source does not state other prices.
What do I need before running it?
You need uv and Python, plus an LLM_API_KEY. Granola, Linear, Gmail credentials and Slack setup are optional. Without them, the skill runs on sample data and never posts to Slack.
Will it send emails or change my Gmail?
No. Gmail access is read only. The skill reads your inbox to build memory, and posts go only to Slack when Slack is set up.
Why does the check say MISSING?
Each MISSING line names one fix needed before the run. The README's What it needs table explains each item. Do not create credentials yourself; fix the named item and run the check again.
Does it work without Slack, Linear or Gmail?
Yes. Slack, Linear and Gmail are optional. Without them the agent uses sample data and prints the steps instead of posting. You can also skip Linear or Gmail with --no-linear or --no-email.
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
This skill suits teams who run many Granola calls and track follow-ups in Linear and Slack, and developers who want a memory-backed agent example to study. It is a poor fit for quick one-off summaries, since the cookbook setup requires Python, uv and several accounts. Users who need the agent to edit or send email should look elsewhere, because Gmail access is read only.
Download the skill