Agent Exercise 2: Design Your Agent Workflow

Exercise Overview

This is the capstone exercise of the workshop. You'll work in teams to design an AI agent for a real workflow from your own team — using everything you learned on Day 2 (prompt engineering) and Day 3 (agent patterns).

🎯 Deliverable

Each team produces an Agent Design Canvas — a one-page document that describes your agent's role, workflow pattern, skills, guardrails, and expected business impact. You'll present this to the group and get feedback.

Step	What you do	Duration
Step 1	Pick your workflow	5 min
Step 2	Choose a workflow pattern	5 min
Step 3	Design the agent (fill the canvas)	25 min
Step 4	Build a prototype skill in Kiro	10 min
Presentations	Each team presents (5 min each)	30 min

Step 1: Pick Your Workflow

Choose a real workflow from your team that you'd like to automate. Use the homework you identified at the end of Day 2, or pick from these ideas:

Workflow	What the agent would do	Pattern
Invoice processing	Extract → validate → flag mismatches → route for approval	Chaining
Fraud monitoring	Analyze from 3 angles → majority vote → case file	Parallelization
Customer complaints	Classify type → route to correct team → draft response	Routing
Credit applications	Assess → auto-decide if small → human review if large	Orchestration
Regulatory updates	Scan new circular → assess impact → notify compliance	Chaining
Monthly reporting	Pull data → generate narrative → create presentation	Chaining
KYC verification	Parse documents → extract fields → check against rules	Chaining
Merchant onboarding	Review application → risk check → compliance check → approve/reject	Orchestration

⚠️ Best choice: Pick a workflow that your team does at least weekly and that takes at least 30 minutes each time. The bigger the time savings, the stronger your business case.

Step 2: Choose a Workflow Pattern

Match your workflow to one of the four patterns from the Day 3 presentation:

Pattern	How it works	Best for
Chaining	Step A → Step B → Step C (sequential)	Linear processes: extract → validate → report
Parallelization	Run 3 analyses simultaneously → combine	Multi-perspective reviews: risk + product + compliance
Routing	Classify first → route to correct workflow	Incoming documents/requests with multiple types
Orchestration	Complex flow with decision points + human review	High-stakes processes with approval thresholds

✅ Decision: Write down your workflow name and chosen pattern before moving to Step 3.

Step 3: Fill the Agent Design Canvas

Use Kiro to help you design the agent. Paste this canvas template and ask Kiro to help you fill it in:

AGENT DESIGN CANVAS — Copy & paste into Kiro

Help me design an AI agent using this canvas template. I'll describe my workflow and you help me fill in each section. # Agent Design Canvas ## Agent Identity - **Name:** [descriptive name for the agent] - **Role:** [Use the persona formula: Title + Experience + Specialty + Characteristic + Behavior] - **Trigger:** [What event starts this agent? e.g., "New file uploaded", "Weekly schedule", "Manual request"] ## Workflow - **Pattern:** [ ] Chaining [ ] Parallelization [ ] Routing [ ] Orchestration - **Steps:** 1. [First action — what skill runs? what does it produce?] 2. [Second action — what does it do with the output of step 1?] 3. [Third action — decision point, human review, or final output?] ## Data - **Inputs:** [What data does the agent need? Files, databases, APIs?] - **Outputs:** [What does it produce? Report, notification, decision?] - **Knowledge base:** [What documents should it reference? Policies, guidelines?] ## Guardrails - **Must NOT:** [List negative constraints — what the agent must never do] - **Escalate when:** [When does a human take over? Thresholds, confidence levels?] - **Data rules:** [Grounding rules — only use provided data? Citation requirements?] ## Kiro Implementation - **Steering file:** [What global rules apply?] - **Skills needed:** [List each SKILL.md file needed for the workflow steps] - **Hook trigger:** [What event triggers the workflow?] - **MCP connections needed:** [What external systems? e.g., database, Slack, email, Google Drive — your tech team sets these up] ## Business Impact - **Current process:** [How is this done today? How long does it take?] - **With agent:** [Expected time savings per week/month] - **Success metric:** [How will you measure if the agent is working?] --- My workflow is: [DESCRIBE YOUR WORKFLOW HERE] The pattern I chose is: [YOUR PATTERN] My team currently spends [TIME] on this each [FREQUENCY].

💡 Tips for a strong canvas:

Be specific about the trigger — "new CSV file in /data" is better than "when needed"
Name each skill — each step should map to a SKILL.md file
Include escalation rules — when does a human take over?
Quantify the business impact — "saves 10 hours/week" is more compelling than "saves time"
Use Day 2 techniques — persona for the role, structured output for each skill, negative constraints for guardrails

Step 4: Build a Prototype Skill

Pick the most important step from your workflow and build it as a working SKILL.md in Kiro:

PROMPT — Build the first skill

Based on the Agent Design Canvas we just created, build the SKILL.md file for Step 1 of the workflow. Create it at .kiro/skills/[agent-name]/SKILL.md with: - Proper frontmatter (name, description with trigger keywords) - The full prompt template for this step - A "When to Use" section - A "Guardrails" section with the constraints from the canvas - An "Output Format" section defining the expected structure Then create a simple hook that triggers this skill when the appropriate file event occurs.

✅ Checkpoint: You should now have:

A completed Agent Design Canvas (the full workflow design)
A working SKILL.md for the first step of your workflow
A hook that triggers it automatically

Presentations

Each team presents their Agent Design Canvas to the group (5 minutes per team):

Cover this	Time
The problem: What workflow are you automating? How long does it take today?	1 min
The design: Which pattern? What are the steps? What skills are needed?	2 min
The guardrails: What must the agent NOT do? When does a human take over?	1 min
The impact: How much time/cost does this save? What's the success metric?	1 min

💡 Feedback questions for the audience:

What could go wrong? Are the guardrails sufficient?
Should the autonomy level be higher or lower?
What data sources would make this more powerful?
Could this pattern apply to a similar workflow in your team?

For Your Team: Kiro ↔ Claude Cowork

Everything you built today is portable. Here's how the concepts translate:

Concept	In Kiro (workshop)	In Claude Cowork (your desk)
Skills	.kiro/skills/*/SKILL.md	SKILL.md in project folder
Always-on rules	.kiro/steering/*.md	CLAUDE.md project instructions
Auto triggers	Hooks (.kiro/hooks/)	Scheduled tasks (built-in)
External tools	MCP servers	Cowork plugins (Slack, email)

🎉 What You've Accomplished (3-Day Workshop)

Day 1: Understood what AI can do and identified use cases for your team
Day 2: Mastered prompt engineering and built reusable templates
Day 3: Converted templates into skills and designed an automated agent workflow
Deliverables: A prompt template, a SKILL.md file, a steering file, a hook, and an Agent Design Canvas — all ready to share with your team

← Exercise 1: Convert to SKILL.md Workshop Home →

🤖 Agent Exercise 2: Design Your Agent Workflow

Exercise Overview

Step 1: Pick Your Workflow

Step 2: Choose a Workflow Pattern

Step 3: Fill the Agent Design Canvas

Step 4: Build a Prototype Skill

Presentations

For Your Team: Kiro ↔ Claude Cowork

🎉 What You've Accomplished (3-Day Workshop)