Wallet Twin Court

Use this skill when the user wants to put a Solana wallet on trial, identify the action most likely to cause regret tomorrow, return a verdict, and only then preview execution.

Use this skill to run the 后悔药 / Wallet Twin Court workflow inside OpenClaw. The job is not to produce many ideas. The job is to put one wallet on trial and compress the result into:

• 你的后悔体质
• 今日被起诉的操作
• 法庭裁决
• 执行前审理
• 判后报告

This skill is designed for OpenClaw x OKX OnchainOS.

Public endpoint:

• Court API: https://todays-orders.vercel.app/api/todays-orders

When to use it

• The user wants a pre-trade judgment for one Solana wallet.
• The user asks what action they are most likely to regret tomorrow.
• The user wants one verdict instead of many trade ideas.
• The user wants one explicit prosecuted action written down.
• The user wants a quote-backed hearing before execution.
• The user wants to execute only after a verdict and then summarize the result with a receipt-backed closing report.

Core law

Every response must obey this rule:

Every action must go through court first. High-regret trades do not get clearance.

Required capabilities

Use OKX OnchainOS as the factual layer:

• Wallet / Portfolio for balances, holdings, stablecoin reserve, concentration, idle capital, and historical mistake context
• Market for price, 24h move, and whether today is a high-temptation regime
• Trade for quote, route comparison, and execution hearing
• Broadcast / Status for signature, confirmation, and receipt

Do not invent holdings, quotes, routes, or receipts.

Do not substitute this skill with miraix-wallet-roast. Do not call https://app.miraix.fun/api/wallet-audit. Do not append the wallet roast share card URL. This skill must use the todays-orders court endpoint and return the courtroom structure, not the roast structure.

Workflow

1. Extract one Solana wallet address. If none is provided, ask for it.
2. Run:

bash

curl -sS -X POST https://todays-orders.vercel.app/api/todays-orders \
  -H 'Content-Type: application/json' \
  -d '{"walletAddress":"<wallet-address>"}'

3. Base the answer on the returned JSON.
4. Read the wallet dossier first:
   • total wallet value
   • top holdings
   • stablecoin reserve
   • concentration
   • idle capital
5. Build the wallet twin's regret profile:
   • what kind of mistake is most likely today
   • whether today is a chase, revenge, over-rotation, or watch-only day
6. Prosecute exactly one action that should not happen today.
7. Return exactly one verdict:
   • approve
   • probation
   • size-cap
   • reject
8. If execution is requested and the verdict is not reject, use the returned executionPreview and executionPlan as the hearing record.
9. Only continue toward execution if the client environment can sign and broadcast safely. Otherwise stop at preview and say execution should continue in the wallet-connected web client.
10. Close with a final report based on the actual state:

• preview report if not executed
• receipt-backed report if executed

Important returned fields:

• todayIntel
• forbiddenOrder
• approvedOrder
• executionPreview
• nightDebrief
• sentenceLadder
• verdictCertificate
• machineView
• executionPlan

Fixed output

Always return the result in this order:

1. 你的后悔体质
   • one-line profile
   • wallet summary
   • 2-3 evidence bullets
2. 今日被起诉的操作
   • one explicit action the court is prosecuting today
3. 法庭裁决
   • exactly one verdict, or explicitly say no action is cleared
4. 执行前审理
   • quote, route, slippage, expected output
   • label clearly as preview until executed
5. 判后报告
   • preview report if not executed
   • receipt-backed report if executed

Output guidance

• Keep the tone disciplined and procedural.
• The shell can be vivid, but the evidence must come from wallet, market, quote, and receipt data.
• Do not output 评分 unless that field truly exists in the court API output.
• Do not output a roast-style 分享卡片 URL.
• Do not rename the sections into wallet roast headings.
• If no clean setup exists, return reject or watch-only and explain why.
• Do not treat simulation as execution.
• If the user asks to execute, ask for final approval before broadcast.