regression-watch

1# Regression Watch
2 
3Detect whether Claude Code sessions are getting worse over time across quality and
4efficiency metrics, using Agent Monitor data.
5 
6## Input
7 
8The user provides: **$ARGUMENTS**
9 
10This may be:
11- empty or "all" — check every regression metric (default)
12- "errors" — error-rate regression only
13- "cache" — cache hit-rate regression only
14- "compaction" — compaction-frequency regression only
15- "cost" — cost-per-session regression only
16- A window like "last 30d" or "30 vs 90" — set the recent vs baseline window sizes
17 
18## Data Sources
19 
20| Endpoint | Returns |
21|----------|---------|
22| `GET /api/analytics` | `daily_events` (365d), `daily_sessions` (365d), `event_types`, `tokens` (total_input, total_output, total_cache_read, total_cache_write — baselines pre-summed), `avg_events_per_session` |
23| `GET /api/events?session_id=X` | Event stream incl. `APIError`, `Compaction`, `PreToolUse`/`PostToolUse` — used to localize regressions to specific sessions |
24| `GET /api/pricing/cost` | `{ total_cost, breakdown[...] }` — total cost to derive cost-per-session |
25| `GET /api/pricing/cost/{sessionId}` | Per-session cost — used to compare recent vs baseline session cost |
26| `GET /api/workflows/{sessionId}` | `compaction` (impact), `errorPropagation` (by depth), `effectiveness` — per-session quality signals |
27| `GET /api/sessions?limit=N` | Sessions with `started_at`, `cost`, `metadata` — to bucket sessions into time windows |
28 
29## Report Sections
30 
31### 1. Windowing
32…