Agent interface — AI Launchpad (read-only)

AI Agent Workbook and Templates

Human site · עברית · llms.txt · llms-full.txt · JSON index · Markdown

← All guides · JSON · Markdown

A hands-on workbook for designing personal, family, and friend-facing agents with reusable templates, canvases, and checklists.

למה זה חשוב: כדי לעבור מתיאוריה לבנייה, צריך templates ו-workbook. החלק הזה הופך את המדריך לכלי עבודה שאפשר לבנות איתו סוכנים אמיתיים לעצמך, למשפחה ולחברים.

Full content

## Why a workbook matters

Guides teach principles. Workbooks help you build.

This guide turns agent design into a repeatable practice using templates you can copy into documents, notes, or AI tools.

## Template 1: Agent Design Canvas

Use this before building any new agent.

### Name
Short and role-based.

### User
Who is this for? You, family, a friend, a client, or a team?

### Mission
What should the agent help accomplish in one sentence?

### When to use
What situations should trigger it?

### Inputs
What does it need to do its job?

### Outputs
What must it return?

### Allowed tools
Only the minimum tools it truly needs.

### Guardrails
What is forbidden? What must always happen?

### Memory boundary
Does it need current task only, working memory, or long-term personal context?

### Human checkpoint
What actions require approval?

## Template 2: Agent Prompt Skeleton

```text
Role: You are a [role].

Mission: Your job is to [mission].

When to use: Use this agent when [trigger conditions].

Inputs: You receive [inputs].

Outputs: Return [required outputs].

Allowed tools: [minimum tools].

Guardrails:
- Do not ...
- Always ...
- Ask for approval before ...

Procedure:
1. Understand the task
2. Check constraints
3. Execute the task
4. Return the result in the required format
```

## Template 3: Security checklist

Before activating an agent:
- Does it have more tools than necessary?
- Does it have access to memory it does not need?
- Can it trigger external actions without approval?
- Can it leak private context into shared spaces?
- Are secrets stored outside prompts and markdown files?

## Template 4: Reality check checklist

Ask:
- Is this really a separate agent, or just a prompt variant?
- Does this agent save time or reduce mistakes?
- Can I describe its value in one sentence?
- Do I know what failure looks like?
- Do I know how I will improve it after the first version?

## Starter agent ideas for yourself

### Personal chief-of-staff
Helps with prioritization, next steps, and focus.

### Research assistant
Finds and compares sources before you decide.

### Writing partner
Turns rough ideas into clearer drafts.

### Memory curator
Maintains continuity across projects and weeks.

## Starter agent ideas for family

### Family logistics helper
Summarizes logistics, reminders, shopping, and schedule coordination.

### Kids learning guide
Helps explain topics simply and gives practice prompts.

### Household planner
Breaks down recurring chores and decision lists.

## Starter agent ideas for friends

### Meeting prep helper
Creates agenda, prep notes, and follow-up points.

### Decision support helper
Compares options and highlights tradeoffs.

### Opportunity scanner
Researches a tool, service, or market trend quickly.

## Suggested weekly practice

Every week:
1. build or refine one agent
2. use it on one real task
3. log one failure mode
4. tighten tools, outputs, or guardrails

This is how agent systems improve in practice.

## Related guides
- Agent architecture
- Practical workflow
- Topologies and diagrams

Human guide page (canonical)

Dror Moshe Aharoni / דרור משה אהרוניAha.Dror.AI@gmail.com WhatsApp

Detected agents are redirected here automatically. Override with ?view=agent or ?view=human.