A股信号追踪

使用这项技能来跟踪现有金融或投资信号的演变情况。当用户提供现有的论点/信号后，可以询问新的市场信息是对信号的支持、削弱、虚假报道、证实还是毫无影响。需要对比基准逻辑与更新后的新闻、价格和基本面数据，然后输出一个结构化的信号更新结果。如有需要，可以使用 mx-search 和 mx-data 来获取最新信息；但不要仅用于简单的新闻查询、原始数据查询、股票筛选或仅查看图表等任务。

AlphaEar Signal Tracker

Track whether new market information changes an existing investment signal.

Inputs

• Required: baseline signal or thesis, plus new event/time window or a request to refresh latest information.
• Preferred baseline fields: title, affected tickers, transmission chain, expected catalyst, timeframe, confidence, sentiment, and falsification conditions.
• If no baseline signal exists, first build one from the user's thesis and label assumptions clearly.

Workflow

1. Parse the baseline signal: thesis, affected tickers, expected mechanism, time horizon, and falsification criteria.
2. Gather fresh facts:
   • Use mx-search for news, announcements, reports, policy, and event context.
   • Use mx-data for price, volume, financial metrics, and other exact data.
3. Compare new information with each baseline logic node.
4. Classify evolution as Strengthened, Weakened, Falsified, Realized, or Unchanged.
5. Update confidence, sentiment, expectation gap, and affected tickers only when evidence supports it.
6. Use references/PROMPTS.md for the research, analysis, and tracking prompt patterns when generating strict signal JSON.

Output Contract

• Start with the evolution label and one-sentence reason.
• Include evidence grouped by news/event, price/data, and logic-chain impact.
• Show what changed from the baseline and what did not.
• Return structured JSON when the downstream workflow needs InvestmentSignal; otherwise provide a compact human-readable update.
• Separate facts from inference and state unresolved uncertainties.

Failure Handling

• Missing baseline: ask for the prior signal or create an explicitly provisional baseline.
• No fresh data: state that evidence is insufficient and do not change the signal.
• Conflicting evidence: mark mixed impact and explain which logic nodes conflict.
• Data/search failures: report the failed source and avoid filling gaps from memory.

Validation

• Run or review tests/test_tracker.py after changing scripts.
• Use evals/evals.json to ensure tracker prompts trigger this skill while simple search/data/diagram prompts route elsewhere.