Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed, non-executable A-share analysis methodology, but its trading recommendations need independent verification before use.
Install only if you want a market-analysis template, not an automated trading authority. Verify the target-price formulas, require dated source citations for any data used, and do not rely on the generated buy/sell or liquidation language without independent financial review.
The target price labels contradict the formulas: the so-called optimistic case uses a lower multiple while the pessimistic case uses a higher multiple. This can invert risk interpretation and produce materially misleading buy/sell guidance, especially because the skill converts these values into concrete trading actions.
The manifest says the skill performs five specific dimensions: technical, valuation, institutional, capital flow, and peer comparison. In the body, the documented pipeline also collects financial fundamentals, business composition, and top-shareholder chip structure, and later permits fallback to other web data sources, which materially broadens the behavior beyond the stated five-dimension diagnosis.
The skill outputs explicit investment ratings, target entry ranges, and stop-loss guidance without a clear warning that the content is not financial advice and may cause losses. In this context, missing safety framing increases the chance that users over-trust automated recommendations and act on them without understanding uncertainty or suitability.
The rating table goes beyond analysis and prescribes concrete actions such as building positions, reducing by 50%, or liquidating, yet provides no explicit risk warning or suitability guardrails. Because these outputs can directly influence asset allocation and liquidation decisions, the omission creates a significant risk of financial harm from erroneous, stale, or oversimplified model results.
Authorizing fallback to unspecified web or other data sources weakens trust boundaries and undermines the stated data provenance of akshare/baostock. In a financial recommendation skill, silently sourcing from arbitrary web content can introduce manipulated, stale, or adversarial data that changes outputs and recommendations.
