Agent Trainer

---
name: agent-trainer
description: Use when you want to teach your AI agent about you. A guided Q&A that builds a personalised profile — saving preferences, style, priorities, and pet peeves directly to memory.
version: 1.0.0
author: Pete Swain / Hermes Agent
output_mode: chat
---

# Agent Trainer

## Overview

Your agent is only as useful as what it knows about you. This skill runs a structured conversation to extract the things that matter — your communication style, how you make decisions, your key relationships, what good looks like, and what drives you up the wall in AI interactions. Everything you share gets saved to the agent's persistent memory so every future session starts smarter.

## When to Use

- When you're setting up an agent for the first time
- When you've been frustrated by the agent getting something wrong repeatedly
- When your situation has changed (new business, new priorities, new role)
- When the agent feels generic and needs personalising
- Trigger phrases: "train you", "teach you about me", "agent training", "let's do an agent session"

## How It Works

The agent works through 7 topic areas, asking one focused question at a time. Take as long or as short as you like on each. If you want to skip a topic, say "skip."

1. **Communication style**
   How do you prefer information delivered? Short and punchy or detailed with context? Do you want pushback or just execution? Any specific things you hate (bullet lists, corporate speak, over-explaining)?

2. **Decision-making style**
   How do you make decisions? Do you want to be talked through them or do you want a recommendation? How risk-tolerant are you? Do you prefer speed or certainty?

3. **Key relationships**
   Who are the important people in your work and life that the agent should know about? Names, roles, and any context that helps (e.g. "Liz is my co-founder and wife — always loop her in on big decisions").

4. **Current priorities**
   What are you focused on right now? What's the #1 thing you're trying to move forward in your business or life in the next 90 days?

5. **What good looks like**
   When has the agent helped you in a way that felt genuinely useful? What made it good? (Or if it's your first time, describe your ideal AI assistant interaction.)

6. **What you hate**
   What behaviours, patterns, or outputs from AI systems drive you crazy? Be specific. This is not about politeness — it's about not repeating mistakes.

7. **Anything else**
   Is there anything about you, your work, or your situation that you want the agent to know that wasn't covered above?

After each answer, the agent:
- Acknowledges what it heard
- Saves the relevant facts to memory
- Confirms: "Saved. Next question..."

## What the Agent Needs From You

Nothing upfront. The skill opens with a brief intro then starts the conversation. Come as you are — rough notes or full paragraphs, both work fine.

## Output

All relevant facts are saved to the agent's persistent memory. The conversation stays in chat. At the end, the agent summarises what it's learned and confirms everything was saved.

No document is produced. The output is a smarter agent in every session that follows.

## Notes

This skill is most valuable run once properly, then revisited every 3-6 months or whenever your situation materially changes. Think of it as an onboarding document that lives in the agent's memory instead of a folder nobody reads.