analyticslisted

# Orchestration Analytics The user wants to see aggregate performance analytics across orchestration history. ## Protocol 1. **Parse arguments**: `$ARGUMENTS` - If a number N is provided (e.g., `last 10`, `10`): limit analysis to the N most recent orchestrations. - If empty: analyze all orchestrations. 2. **Scan history via MCP:** Call `mcp__orchestray__history_query_events` with: - `event_types: ["orchestration_start", "orchestration_complete", "agent_stop", "task_completed_metrics", "routing_outcome", "replan", "verify_fix_start", "verify_fix_pass", "verify_fix_fail", "verify_fix_oscillation", "verify_fix_attempt"]` - `limit: 500` (the server-side maximum) - If N was specified, filter the returned `events` array client-side to the last N unique `orchestration_id` values (the server does not expose a per-orchestration limit — filter after the call returns). If the call returns `isError: true` or the `events` array is empty, report "No orchestration history found. Run some orchestrations first." and stop. **Fallback (MCP unavailable):** if the MCP call itself fails with a transport error (not an empty result), fall back to the pre-v2.0.11 behavior: Glob `.orchestray/history/` directories, read each events.jsonl, normalize the `event`/`type` field. 3. **Apply limit**: (unchanged — operates on the client-filtered list from step 2.) 4. **Collect data**: The MCP call already returns the normalized event objects. No JSONL parsi