Security audit
Security checks for vulnerabilities and agentic risk
The skill is a stock-sentiment helper, but it asks for broad command execution and has an overbroad trigger while the referenced scanning script is not actually present in the artifact.
Review this before installing because it grants exec access for a tool whose actual scanner script is missing from the package. Use it only for explicit stock-sentiment requests, avoid entering private watchlists or proprietary identifiers unless you accept external lookups, and do not let the agent run arbitrary commands beyond a known scanner script.
The trigger condition is so broad—covering generic data analysis, report generation, statistical insight, and visualization—that the skill may be invoked for many unrelated requests. In an agent environment with exec and network-enabled workflows, over-triggering increases the chance of unnecessary script execution, external data access, and user confusion about what actions will be taken.
The display name and primary description are presented entirely in Chinese, and the document does not indicate that users can choose their preferred language. For a generally applicable investment-analysis skill, this can violate language/locale policy unless the locale restriction is explicitly justified or made opt-in.
The description repeatedly frames the skill as a lightweight sentiment scanner for one stock, but the manifest exposes the generic exec tool rather than a narrowly bounded scanning interface. While script execution may be an implementation detail, granting unrestricted command execution is materially broader than the behavior implied by a simple investor-facing analysis tool.
The skill describes scanning announcements, news, reports, and social media, but it does not present an upfront warning in the main description that using the skill will access external network resources and potentially transmit query terms to third-party data sources. In an agent setting, this can lead to unanticipated data disclosure, especially if a user includes sensitive watchlists, identifiers, or proprietary symbols in prompts.
The manifest describes standardized sentiment scoring and event listing for a single stock, but the best-practice commands add file output and persistent logging behavior not mentioned in the skill description. This is a broader operational behavior than a simple read-and-report analysis flow, even though it is presented as normal usage.
No suspicious patterns detected.
