| name | Classify Activity |
| description | Group raw activity signals into thematic clusters with effort estimates. |
Input
Raw signals from Work IQ + GitHub for the day.
Output (JSON)
{
"groups": [{
"theme": "auth migration",
"activities": [{ "description": "Reviewed token refresh PR #1234", "source_breadcrumb": "github.com/org/repo/pull/1234" }],
"effort": "high"
}]
}
Instructions
- Scan all signals from the day.
- Cluster by work theme (e.g., "auth migration", "team coordination", "code review").
- Within each cluster, write one sentence per activity describing what was done.
- Every activity MUST have a
source_breadcrumb — a link, thread subject, or channel name. If unavailable: "source unavailable".
- Teams meeting links (
19:meeting_...@thread.v2): these often break after the meeting ends. When a breadcrumb contains a meeting thread link, add a fallback — include the meeting title, date, and participant names so the user can locate the context manually. Mark the breadcrumb with ⚠️ to indicate it may not resolve.
- Estimate effort per cluster: "high", "medium", or "low" relative to the day.
- Max 15 activities total across all groups.