Operations Manual.

OPERATIONS MANUAL

Lighthouse Sprint — Stage 4: Handoff

Your team's complete guide to running the agentic workflow day-to-day.

Provide a plain-language summary of what the agentic workflow does. Write this for the operators who will run it daily — not for engineers.

1.1 What This System Does

[Describe the workflow in 2-3 sentences. What goes in, what comes out, who benefits.]

1.2 Workflow Architecture

Paste or reference the workflow diagram from the Redesign Blueprint. Label every agent, human checkpoint, and data flow.

[Insert workflow diagram or reference to diagram file]

1.3 Agents in This Workflow

List every agent deployed in this workflow. One row per agent.

What does an operator need to do every day to keep this workflow running?

2.1 Morning Checklist

List the checks an operator should run at the start of each day.

[1. Check monitoring dashboard for overnight alerts]

[2. Review agent queue — any stuck or failed items?]

[3. Verify data inputs are flowing — check source systems]

[4. Review any items flagged for human approval]

[5. Check governance dashboard — any escalations pending?]

2.2 During the Day

What ongoing tasks does the operator manage?

[Describe the steady-state operation — what the operator monitors, when they intervene, how they handle the human-approval queue.]

2.3 End of Day

What wrap-up activities are needed?

[1. Review daily throughput — compare to baseline targets]

[2. Clear any remaining approval queue items]

[3. Note any anomalies in the daily log]

[4. Confirm all agents completed their scheduled runs]

2.4 Weekly Tasks

[List any weekly maintenance, reporting, or review tasks.]

For every agent-initiated, human-approved task — what does the operator need to review, and how do they approve or reject?

3.1 Approval Queue

[Describe where approval requests appear — dashboard, email, Slack, etc. How does the operator access them?]

3.2 Review Criteria

For each type of approval, what should the operator check before approving?

3.3 Handling Rejections

[What happens when an operator rejects an agent's output? Does it retry, escalate, or require manual completion?]

Common issues and their solutions. Written for operators, not engineers.

4.1 Common Issues

4.2 Escalation Path

[Level 1: Operator tries the fix from the table above]

[Level 2: Contact [Team Lead / Technical Support] via [channel]]

[Level 3: Contact Umbra Studio support (within 30-day window) via [channel]]

5.1 Key Metrics to Track

5.2 Monitoring Dashboard

[URL or access instructions for the monitoring dashboard]

[What to look at, how often, what normal looks like]

Track all updates to this manual.