---
id: owliver-conversation-test
name: Workspace Activity Assistant
description: Answer questions about applications across the workspace and the roles still open.
pages:
  - positions
status: active
triggers:
  - workspace applications
  - applications across the workspace
  - roles still open
  - roles needing attention
owliver:
  enabled: true
  # Every suggestion names the capability it asks for, so a chip and an answer
  # are one-to-one rather than both falling through to the first declared
  # capability.
  suggestions:
    - label: Summarize workspace applications
      capability: summary
    - label: List the roles still open
      capability: list
  capabilities:
    - summary
    - list
  responses:
    summary:
      title: Workspace Applications
      source: candidates.activity
      periods:
        - today
        - last-7-days
        - previous-month
    list:
      title: Roles Still Open
      source: positions.demand
      limit: 5
---

# Workspace Activity Assistant

## Purpose

The panel half, and only that half. No `ui:` block, so nothing is drawn on the
Positions page and no card appears above or inside the grid.

Each response names its own `source:` directly, which is what a definition
without a `ui:` section must do — `normalizeSkillOwliver` inherits a source from
the first `ui:` section only when one exists, and reports
`owliver.responses.<capability>: a response needs a source, or a ui: section to
read from` when neither is present.

Both sources declare `context: null`, so neither answer asks which position is
meant: `resolveEntity` returns immediately and no
"This section needs a position to read" is possible. The shapes match what each
source offers — `summary` is prose and is exempt from the shape check, and
`list` is a shape `positions.demand` declares.

## Capabilities

- Summarize applications across the workspace over today, the last 7 days and
  the previous month.
- List the open positions and how many applicants each has.