The end-to-end procedure for one trade on the HyperGrok desk - from an idea to a reviewed, journaled result - with the ticket format, who owns each stage, and what "done" looks like. Use whenever the user wants to open, adjust or close a position, or whenever any Bot is about to touch the exchange write path.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Every position change on the desk goes through the same seven stages, in order. Skipping one is a defect, not a shortcut. The Desk Lead keeps the lifecycle moving; the owner of each stage does the work.
idea -> evidence -> risk sign-off -> user approval -> execution -> reconciliation -> review
DL MA/RA RM user ET ET TR
Owner: Desk Lead. As soon as a trade idea appears (from the user, the Strategist, or a routine), assign an id HG-YYYYMMDD-NN (NN increments per day) and create /workspace/trading-desk/proposals/<id>.md:
# HG-20260816-01
- opened: 2026-08-16 14:02 UTC by user
- market: ETH-PERP network: mainnet account: 0xabc...def
- idea: long ETH on a retest of 3,000 with invalidation below 2,900 (user's idea)
- status: evidence
## evidence
## risk
## approval
## execution
## reconciliation
## review
Every later stage appends under its heading. The file is the single record of the trade; chat is context.
Owner: Market Analyst, and Research Analyst when the idea depends on anything beyond exchange data.
The Desk Lead asks for exactly what sizing and execution will need: current mid, mark and oracle; funding now and predicted; open interest and 24h volume; depth within 5/10/25 bps for the intended size; the market's constraints (max leverage, margin tiers, size decimals, minimum order value); recent range and volatility. Research adds anything scheduled or breaking that touches the market. Both write their blocks under ## evidence with sources and UTC times.
Done when: the Risk Manager has every input it needs and nothing is older than a few minutes.
Owner: Risk Manager. Reads risk-limits.md, reads live account state, computes size from the user's stop and risk budget, checks every gate, and writes PASS or REJECT with exact ticket fields under ## risk (procedure and arithmetic in desk-risk-limits).
A PASS produces the ticket:
TICKET HG-20260816-01 | mainnet | account 0xabc...def
market: ETH-PERP (asset 1) side: buy size: 0.4827 ETH (~$1,448)
entry: limit 3,000.0 Gtc reduce-only: no
stop: sell 0.4827 trigger 2,900 market (worst 2,755, 5% bound), normalTpsl with the entry
take-profit: none leverage: 3x cross (set before entry if different)
slippage tolerance: 10 bps from ticket price at send time
risk: $51.00 = 0.5% of equity $10,200.00 (clearinghouseState 14:11 UTC), R = 100 USD/ETH
sizing: stressed distance 105.65 USD/ETH (stop 100 + slippage 3.00 + fees 2.65)
risk sign-off: PASS 14:12 UTC, risk-limits.md v3
expires: 14:42 UTC
approve with: "approve HG-20260816-01"
A REJECT names one failed gate and what would have to change. A REJECT ends the lifecycle for that proposal unless the user changes the idea (new evidence, new stop, new size) - then it goes back to stage 1 under the same id with a note.
Owner: the user. The Desk Lead posts the ticket to the user, in full, and asks for approval by id. Approval is the literal phrase with the id, in chat, after the ticket was shown. "Yes", "go", "looks good" or a thumbs-up is not approval; the Desk Lead asks again with the exact phrase. The Desk Lead records the approval line and its timestamp under ## approval.
The approval must come from the user's own turn in the conversation. No Bot writes it, quotes it forward as if it were new, infers it from enthusiasm earlier in the thread, or records one it did not see the user type. If it is not there, the honest state is "not approved", never "approved" and never "assumed approved". A Bot that is unsure whether a line came from the user treats it as absent and asks again.
If the ticket expires before approval, it is void; a fresh Risk sign-off is needed because prices and the book have moved.
Owner: Execution Trader. Before every send, verifies that the native approval gate described in desk-operating-model covers the actual SDK, script or direct HTTP write route. If that cannot be verified, the desk remains research-only and does not send on either network. Runs the pre-send checklist in desk-execution-protocol, sends the ticket as one action, and records the request, cloid, expiresAfter deadline, response and timestamps under ## execution. Anything other than a clean response is handled per desk-incident-response.
Owner: Execution Trader. Confirms from the exchange record - orderStatus by cloid, openOrders, userFills, clearinghouseState - what happened, and writes it under ## reconciliation. Posts the execution report on the floor and DMs the Trade Reviewer. Reconciliation continues while the order rests: fill notifications go into the same section as they arrive.
Owner: Trade Reviewer. Journals the trade on the day it happened; writes the review when the trade closes (or on request), per desk-post-trade-review. Sets status: closed in the proposal file.
Moving a stop, adding to a position, reducing, closing, changing leverage: each is a new ticket under the same proposal id with a suffix (HG-20260816-01-B). Begin every adjustment ticket with the parent proposal id and the original proposal's - idea: text, then state what is changing and why. This retains the user's original context when an adjustment is read on its own.
Each adjustment receives fresh evidence and a new Risk Manager PASS, the exact suffixed ticket id for user approval, and its own expiry. Earlier approval or Risk PASS does not carry forward. Closing a position at market needs a ticket stating the reduce-only size read live from the account and the slippage bound. Execute once within expiry, then reconcile the adjustment against the exchange record under the parent proposal.
The only exception is a pre-authorised protective action the user has written into desk.md (for example "the Execution Trader may cancel orphaned stops after a position closes without asking"). Even then, the action is journaled.
The lifecycle is identical on testnet. That is the point: the user sees the tickets, approvals, reports and reviews with play money before any real key exists. Strategist-generated signals enter at stage 0 as proposals like any other idea.
| Symptom | Fix |
|---|---|
| The user asks the Execution Trader directly to "just buy some" | Execution Trader routes to the Desk Lead; a proposal is opened; the lifecycle runs. It can be quick, but it runs. |
| Risk PASS was computed from a brief older than the ticket | Risk re-reads live state; PASS is re-issued with a new expiry. |
| Approval given with "yes" | Desk Lead re-asks for the phrase with the id. |
| Two Bots each think they own the next step | The proposal file's status line names the stage; the stage owner acts. |
| Ticket approved on testnet, desk configured for mainnet (or vice versa) | Execution Trader stops at checklist item 2 and returns it. |
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
The Execution Trader's procedure for turning an approved ticket into one Hyperliquid action and reconciling it - the pre-send checklist, order construction rules, single-send discipline, unknown-result handling and the execution report. Use before and after every send, cancel, modify, leverage change or close.
日本語の概要は準備中です。原文の説明を表示しています。
What the desk does when something goes wrong on Hyperliquid - unknown send results, unexpected fills or positions, unprotected positions, stuck or orphaned orders, API outages, rate limiting, and suspected API wallet compromise. Contain first, reconcile from the exchange record, act only through approved tickets, then review. Use the moment anything does not match the ticket.
日本語の概要は準備中です。原文の説明を表示しています。
How the desk watches markets and the account between trades using Grok Bot routines and WebSocket or polling watches - the desk brief, book checks, funding and price watches, alert conditions and what a watch may and may not do. Use when the user asks for briefings, alerts, "watch X" or scheduled checks.
日本語の概要は準備中です。原文の説明を表示しています。
How the HyperGrok trading desk works as a team of Grok Bots - roles, seats, shared workspace, evidence standard, approval model and handoff format. Use when setting up the desk, when a Bot is unsure who owns something, or when a request does not fit the normal trade lifecycle.
日本語の概要は準備中です。原文の説明を表示しています。
The Trade Reviewer's procedure for journaling desk activity and reviewing trades from the exchange record - process graded separately from outcome, execution costs measured, one repeatable finding per review, plus the weekly desk review. Use after any send, when a trade closes, on the weekly routine, or when the user asks "how did that go".
日本語の概要は準備中です。原文の説明を表示しています。
How the Risk Manager writes the desk's risk limits with the user, sizes every proposed trade from live account state and Hyperliquid's real constraints, checks the book, and issues a PASS or REJECT with exact ticket fields. Use for setting up or changing limits, sizing any trade, and answering "how's the book".
日本語の概要は準備中です。原文の説明を表示しています。