AOP: {Procedure Name}

Purpose

{What does this agent procedure accomplish? What would require human effort without this AOP?}

Trigger

• When: {What triggers this AOP? Schedule, event, human command, or another AOP}
• Frequency: {one-time | on-demand | daily | weekly | on-event}
• Autonomy: {L0: human runs it | L1: agent suggests, human approves | L2: agent runs, human reviews | L3: fully autonomous}

Agent Assignment

• Primary agent: {agent name — e.g., Legacy, Spartan, Oracle}
• Fallback agent: {backup agent if primary unavailable}
• Human escalation: {when should the agent stop and ask a human?}

Inputs

Input	Source	Required
{input 1}	{where the agent gets it — file path, API, user prompt}	yes/no
{input 2}	{source}	yes/no

Steps

1. {Step 1} — {clear instruction an agent can follow without interpretation}

   • Tool: {specific tool or API to use}
   • Expected output: {what this step produces}

2. {Step 2}

   • Tool: {tool}
   • Expected output: {output}

3. {Step 3}

   • Tool: {tool}
   • Expected output: {output}

Outputs

Output	Format	Destination
{output 1}	{markdown file, JSON, notification, etc.}	{where it goes — file path, inbox, dashboard}
{output 2}	{format}	{destination}

Success Criteria

• {How does the agent (or reviewer) know this completed correctly?}
• {What quality checks should be run on the output?}

Error Handling

Error Condition	Action
{input missing or invalid}	{what to do — retry, skip, escalate to human}
{API/tool failure}	{fallback behavior}
{output doesn’t meet quality criteria}	{retry with different approach or escalate}

Guardrails

• {What is the agent NOT allowed to do during this procedure?}
• {What files/domains are off-limits?}
• {Maximum resources (API calls, tokens, time) this AOP can consume}

Last Verified

{YYYY-MM-DD} — {who confirmed this AOP still works correctly}

---