WEEX Monitor Skill

Read manifest.json for routing rules. Open file-index.json only when you need file-level guidance.

This skill owns WEEX position-PnL automated monitor orchestration. It converts a constrained PnL monitor intent into a task DSL, renders localized monitor confirmation text with a confirmation token, requires explicit monitor confirmation, stores local task state in SQLite, appends audit events, evaluates local PnL triggers, can activate and start a bounded account PnL loop from one combined user confirmation, can run account PnL checks or a bounded account loop through weex-trader-skill, and reports results to the current thread.

This skill does not own API credentials, profile storage, vault unlock, REST signing, or direct REST submission. Those capabilities stay in weex-trader-skill. Account monitor commands are orchestration wrappers around weex-trader-skill guard flows and require internal execution authorization that matches the task trading_mode. Do not use internal flag names in user-facing wording; tell the user plainly whether they are authorizing 真实盘 access and real order execution or 模拟盘 access and demo order execution. Use localized trading-mode labels, not environment labels such as raw live or demo and not account labels, in user-facing dialogue and confirmation text. For Chinese, use 模拟盘 and 真实盘; for English, use demo trading and real trading. Every Chinese user-visible monitor confirmation or monitor result report that references private account state, a monitor action, or order execution authority must start with the exact first-line prefix 当前交易环境： followed by 模拟盘 or 真实盘; do not replace it with 盘别: or move it below the title. Keep raw live or demo only for internal task fields and command flags. A single user confirmation may authorize monitor activation and execution only when the confirmation text includes the matched real-trading position or demo-trading position, close action, final task details, trading mode, and mode/order authority; PnL account loops must also include a finite live duration. User-facing monitor confirmation must ask for one simple localized reply word: 确认 for Chinese copy or confirm for English copy; do not ask for longer localized phrases such as "confirm start monitoring."

AI natural-language parsing

When the user gives a natural-language monitor request, the AI layer of this skill must convert the user's monitor instruction into the Task DSL before calling the deterministic script. The script intentionally accepts JSON only; do not pass raw chat text to the script and treat a missing DSL field as a clarification need, not as permission to infer a risky value.

Required extraction fields: