Operator cockpit for a systematic crypto-alpha desk.

01 / setup

Context, scope, and the constraint that bound the design.

VENUE

TradingView widget (BTCUSDT.P · MEXC perpetuals)

DESK

systematic crypto-alpha · 4 wired strategies

EXECUTION

manual only · no algo · no auto-submit

GATE CONTRACT

5 gates · 4 PASS · 1 PARTIAL (ship-acceptable)

NDA

clean by construction · no employer names · no internal paths

02 / method

Pattern → build → gate. Every cycle, the same 5-beat cadence.

Four design beats, each a constraint that shaped a different part of the surface. No beat is optional; the order is the cheapest sequence that converges.

1. 01

   Surface the constraint, not the chrome.

   A cockpit with 12 widgets is a cockpit nobody uses. The first move is the operator's question: "where do I enter?" Everything else (chart, registry, timeframes) is in service of that one question.

2. 02

   Ship under a gate contract, not a vibe.

   Five gates (build · a11y · quality · graph · chrome). Each gate is a binary pass / fail. PARTIAL is acceptable only with a named precedent in the mission ledger. No gate skipped.

3. 03

   Static widget, dynamic signal.

   The TradingView widget is the only iframe on the surface. CSP is tightened to a single host (frame-src https://s.tradingview.com). Everything else is server-rendered HTML + a tiny parser-driven JSON contract.

4. 04

   Empty state is a feature.

   The entry finder ships empty by default. role="status" + aria-live="polite" so screen readers narrate the load. The first fire is a single, deterministic, audit-able event.

03 / result

Six mono-prefixed metrics. Gate verdict table. Live risk register.

Every metric below binds to a mono prefix (>, ≥, =, <, !) so the reader can audit at a glance whether the number is a floor, a ceiling, an exact claim, or a guard.

• >

  gates passed

  4 / 5

  gate 4 PARTIAL ship-acceptable

• ≥

  strategies wired

  4

  registry rows live, signal hook ready

• ≥

  timeframes

  7

  1m / 5m / 15m / 1h / 4h / 1d / 1w

• =

  chart reload on switch

  true

  fresh frameElementId per switch, no stale state

• <

  registry TTL

  5 min

  TTL + mtime invalidate, never stale

• !

  hard guard

  no auto-submit

  assert "submit_order" not in dir() at parse time

5-gate verdict

Risk register (post-ship)

• R1bars data path pending : graceful 200 with signal_value: null until data population.
• R2script-src single-host carve (TradingView) : acknowledged; SRI follow-up RFC.
• R3CSP-block heuristic : offsetHeight === 0 + 4s watchdog; no silent failures.
• R4upstream_error info-disclosure : internal-only; redact before any external surface.
• R5seed_graph.py --verify re-introduces 2 orphan skill nodes : soft regression owned by seed script.

04 / shipped

Endpoints, parser, components. Every artifact is on disk.

Eight artifacts shipped. Each has a kind (endpoint · parser · component), a label, and a one-line detail. The hard guard is in the parser, not the UI: if any strategy file ships a submit_order callable, the registry refuses to load.

• endpoint

  /api/trading/strategies

  strategy registry feed · 4 rows · status + signal + G-progress + confidence + CTA

• endpoint

  /api/trading/entries

  normalized entry-ticket schema · empty-state-aware · deep-linkable

• endpoint

  /api/trading/strategies/{id}/signal

  per-strategy signal poll · graceful 200 with signal_value: null until bars land

• parser

  build/strategy/strategy_registry.py

  573 lines · 5-min TTL + mtime invalidate · hard guard assert "submit_order" not in dir()

• component

  TradingView widget

  MEXC BTCUSDT.P · dark theme · watchlist BTC / ETH / SOL · allow_symbol_change=true · CSP single-host

• component

  7-timeframe switcher

  1m / 5m / 15m / 1h / 4h / 1d / 1w · default 1d · fresh frameElementId on switch

• component

  strategy registry panel

  P10 · status chips · signal · G-progress · confidence · Open MEXC CTA

• component

  entry finder panel

  P11 · empty-state-aware · role="status" · aria-live="polite" · ready for first fire

Handoff + follow-ups

Two follow-ups are filed, both owned: a data-engineer mission to populate the bars path so the entry finder fires its first signal, and a graph-ratchet mission to repair the 2-component fragmentation surfaced at gate 4. Neither blocks ship. Both are in the mission ledger under money-by-next-week · phase-e.