# Sanitized Template — Agent Workflow

Use this skeleton when the original workflow includes proprietary role names, internal tool aliases, private connectors, or internal infrastructure references.

## Remove

- custom agent names and internal role labels
- private tool names and connector aliases
- repository names, cloud paths, queue names, or internal endpoints
- customer names, account IDs, or credentials

## Keep

- the roles in generic form
- the handoffs between roles
- the checks each role performs
- the approval and recovery points
- the route breakpoints where the flow can fail

## Copy-ready skeleton

```text
Objective:
The workflow is meant to produce [target result] for [target buyer or user].

Workflow roles:
- Scout / intake role: finds or receives the task.
- Execution role: performs the main work.
- Review role: checks output quality or compliance with instructions.
- Delivery role: submits or hands off the result.

Workflow sequence:
1. Intake role receives [generic input].
2. Execution role produces [intermediate or final output].
3. Review role checks [quality / correctness / acceptance criteria].
4. Delivery role sends [final output] through [generic route].

Handoffs:
- Intake -> Execution
- Execution -> Review
- Review -> Delivery

Approvals and gates:
- [operator approval / buyer acceptance / platform review / payment release / unknown]

Recovery points:
- If [condition], retry.
- If [condition], stop and escalate.

Evidence available:
- [trace or log / delivery sample / working connector / none]

Known risks:
- [coordination failure]
- [unclear acceptance criteria]
- [route continuity break]
- [hidden human intervention]
```

## Reminder

Describe the structure plainly. A sanitized workflow should still show who does what, where the handoffs are, and where the route can break.