Show a compact completed-agent handoff without reopening the full timeline #30

Parent: #21
Related: #17, #25
Sources: block/buzz#2932 and block/buzz#2941

Why this is valuable

A requester should see what an agent delivered, what was verified, and what remains blocked without reconstructing the answer from a long room or run timeline.

Bounded v1

• Reuse existing persisted task, agent-run, evidence, output, verification, and status data; do not invent a second job protocol.
• Render one compact completion card in Tasks and the related room timeline.
• Show disposition, concise outcome, requested outcome, verification summary, bounded authorised artifact/evidence links, blocker, completion time, and View full timeline.
• Fall back to the existing generic run result when legacy/malformed/incomplete data cannot satisfy the card contract.
• Treat agent output as untrusted evidence and sanitise all text/links.

Acceptance criteria

• Completed, partially completed, failed, cancelled, and blocked runs have truthful distinct summaries.
• The card never claims verification without persisted verification evidence.
• Artifact/evidence links are organisation-scoped, authorised, bounded, and safe.
• Malformed, oversized, stale, or mismatched result data falls back safely.
• Full append-only timeline remains available from one clear action.
• Unit tests cover disposition reduction, bounds, sanitisation, fallback, and tenant scope.
• Browser tests cover successful, blocked, failed, and unauthorised synthetic runs.

Effort guardrail

Presentation/projection slice over existing records. Any missing durable result schema becomes a separate prerequisite rather than hidden scope.