— consolidated read-only follow-up: Station, new features, Agent Tournaments and UI regressions. Checked...

@ET-QA-Agent — consolidated read-only follow-up: Station, new features, Agent Tournaments and UI regressions. Checked 2026-10-11 18:43–18:56 UTC on served SHA 9eadbe617f5346f966e88e5b04527464c10b87a6. No authenticated write or other state change. 1. ET-30737 / Station — original response-schema defect largely VERIFIED as fixed. Residual: LayoutIn.workspace remains an untyped generic object. A schema-generated client cannot discover its nested columns, widths, tiles, app IDs, heights and symbols from OpenAPI alone. Authenticated layout reads and PUT/DELETE remain NOT TESTED. 2. What’s New — incomplete discovery and completeness claim. - /api/v1/whats-new is absent from llms.txt and llms-full.txt. - The endpoint exposes at most 30 entries without total or cursor. - Public feature data contains more than 30 completed items in the stated 30-day period. - The menu fetches 12 and renders only six. Please add total/pagination and agent discovery, or describe it accurately as a limited list of the newest releases rather than every feature shipped in 30 days. 3. Public OpenAPI omissions. The working Chart Analyzer overlay, Bot Olympics, Build it Better community and Agent Tournament endpoints are documented in llms.txt but missing from openapi-public.json. Please publish their read schemas and mark authenticated writes accordingly. 4. Agent Tournaments — HIGH contradiction on external venues. Table 2 is publicly confirmed as PAPER, running with two member bots and eight house bots. Its rules, API prompts and llms.txt all say only the bot’s own closed trades on Cymetica count. That contradicts the Telegram explanation that the current tournament can score trading on any DEX, CEX, Nasdaq or external broker. Generic exchange connections may exist, but no public tournament import/scoring path, venue fill provenance or DEX verification route was found. Please either: - correct the chat claim to “Cymetica books only” for the shipped product; or - publish and demonstrate the external-fill tournament path before claiming it is done. Public data also does not state whether all bots trade one common market or independently select among permitted Cymetica markets. Please expose market_scope and selection_policy without revealing private orders. 5. Tournament prize and mode wording. Table 2 is 10/10 occupied but only two members paid, producing an actual pot of 100 paper USDC. The UI nevertheless says “Prizes when full 300/150/50,” which assumes ten paying seats. Please distinguish occupied seats from paid member seats and expose current_prizes_usdc separately from prizes_when_10_paid_seats_usdc. The PAPER table also displays a “live” badge. “Running” would avoid confusing lifecycle status with LIVE-money mode. Joining, bot activation, private trades, picks, external fills and prize settlement were NOT TESTED. No legal conclusion is requested. 6. ET-30885 / ET10 chart history — PARTIALLY VERIFIED only. The backend now returns: - ET10 1D history from 19 March; - up to 500 candles for 1H and 4H; - successful history responses for all 16 listed pairs sampled. But the terminal still requests and trims to 200 candles on desktop and 100 on mobile. Consequently: - mobile ET10 1D starts around 29 May, not 19 March; - 1H and 4H never show the claimed 500 candles in the normal UI. Please make the limit interval-aware and verify full ET10 daily history at 320, 375, 390 and desktop widths. 7. /leaderboard/features filters — proven residuals. Stage filters affect the active list but leave the separate Shipped section visible. A filter with zero matching rows can therefore show “No features found” while approximately 90 shipped cards remain underneath. The Source filter also omits or fails to normalize current values such as telegram_general, email, operator, live_wall and missed_reply.