Saving & Rating Workflows
Saving a task to the workflow library does not mean creating a separate export file, but rather saving task evaluation and workflow content together to promote it to a reusable example.
Information Saved
Evaluation Metadata
ratinguseCaseworkflowTagsoutcomeNoteapproachNotereuseWhenwatchoutsevaluatedAt
Workflow Content
The server storage record also includes:
workflowSnapshotworkflowContextsearchText
searchText is not a separate form field but is saved based on the search string from the snapshot.
Save Paths
Web
Saved through TaskEvaluatePanel in the Evaluate tab of EventInspector.
The web path is not just a simple note input screen.
- Receives evaluation input values.
- Auto-generates snapshot/context from timeline.
- User can preview or edit fields to review and modify generated values.
- Can revert to auto-generated values with
Regenerate. - On save, sends evaluation +
workflowSnapshot+workflowContexttoPOST /api/tasks/:id/evaluate.
MCP / Agent
Manual saving is also possible via the monitor_evaluate_task tool. Values directly passed in this path are primarily evaluation metadata.
ratinguseCaseworkflowTagsoutcomeNoteapproachNotereuseWhenwatchouts
If snapshot/context override is not sent, the server auto-generates and saves based on the current task timeline.
good and skip
good: Work worth revisiting next timeskip: Kept in library but not prioritized in recommended examples
Sorting generally has good first, with latest evaluatedAt coming first within that.
Where It Is Used After Saving
GET /api/workflowslist panelGET /api/workflows/similarsimilar example searchGET /api/workflows/:id/contentsnapshot/context detail view- Agent's workflow library search rules
Criteria for Good Evaluation Notes
useCaseshould be short and categorizable.outcomeNoteshould immediately show what was resolved.approachNoteshould explain why this approach worked.reuseWhenshould indicate when to revisit next time.watchoutsshould be specific to avoid repeating the same mistakes.