# My Agent Workflow Canvas

| Field | Your answer |
|---|---|
| Outcome | What useful result must exist when this finishes? |
| Recipient | Who will use it, and for which decision or next action? |
| Trigger | A person starts it, a schedule runs it, or a specific event occurs? |
| Inputs | Which documents, records, messages, URLs, or datasets are required? |
| Source authority | Which source wins when two disagree? What makes a source current enough? |
| Working context | Audience, terminology, project state, examples, constraints, style, and prior decisions |
| Output | Format, required fields, destination, naming convention, and evidence links |
| Allowed actions | Exactly what may the agent read, create, or change? |
| Human checkpoints | Which decisions require a named person, and what must that person inspect? |
| Acceptance criteria | Observable conditions that distinguish useful from incomplete or wrong |
| Failure handling | What happens if data is missing, a tool fails, a fact conflicts, or access is denied? |
| Cost and time limits | Requested limits, limits actually enforced by the platform, and who receives alerts |
| Finish line | What proves the work is complete? |
| Owner | Who maintains the workflow, reviews failures, and can stop it? |
| Next review | When will its performance and access be revisited? |

**One-sentence brief:**  
When ______ happens, use ______ to prepare ______ for ______, following ______, and pause for ______ before ______.