Tp2
- Category
- MCP Tool Poisoning
- Confidence
- 85% confidence
- Finding
Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.
- Content
Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly disclosed, but it asks for persistent remote-service credentials and includes broader financial/history access than its TradingView indicator purpose clearly justifies.
Review before installing. Only use this skill if you are comfortable sending financial-analysis prompts to Prana, storing a persistent Prana API key in OpenClaw, and exposing a browser-openable skill history link. Avoid sending personal, account, brokerage, or credential data through the skill.
Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.
文件整体以中文命名和指令示例为中心,示例请求也仅给出中文表达,未说明是否支持其他语言或允许用户选择输出语言。按照语言/区域政策,若技能默认或隐含强制特定语言而没有用户选择机制,属于自然语言策略风险。
L018-L019 将触发条件描述为“当客户描述需要使用查询TradingView 技术指标相关的分析与计算结果时”并给出“帮我分析…”这类日常表达示例,但没有限定明确触发短语、边界或排除条件。这类表述容易与普通聊天中的泛化分析请求重叠,导致技能被意外调用。
The skill retrieves an API key from a remote endpoint without authentication and instructs storing it as a global persistent environment variable, then uses it for subsequent remote agent calls. This creates a serious trust-boundary problem: a remotely supplied credential is persisted locally and may grant ongoing access beyond the immediate user action, while the document itself admits the endpoint is unauthenticated and effectively auto-creates a user/token relationship.
接口调用命令:
curl -sS "https://www.prana.chat/api/v2/api-keys"
接口正常响应示例:
The skill claims to perform TradingView technical-indicator analysis, but it also exposes a function to retrieve a browser-openable purchase/history URL unrelated to that purpose. This expands the skill's effective data-access scope beyond user-expected analysis behavior and can leak account-linked activity metadata or facilitate unauthorized viewing of transaction/history records.
Access to skill purchase/history records is not justified by the stated business purpose of technical indicator analysis. When a skill can retrieve account-related history artifacts without tight scoping, it creates unnecessary exposure of sensitive usage records and increases the chance of privacy violations or confused-deputy behavior.
The skill claims to perform TradingView technical-indicator analysis, but it also declares access to a purchase-history URL endpoint unrelated to that purpose. This expands the data-access surface beyond user-expected functionality and could expose account or commercial metadata without a clear need, violating least-privilege expectations.
The usage examples broaden the skill from technical-indicator analysis to general stock returns, price movement, and financial-report retrieval, which does not match the declared scope. This mismatch can cause over-triggering and lead users or orchestrators to send broader financial queries and data to a remote agent than intended.
Broad usage examples create an ambiguous activation scope, making the skill eligible for general stock and financial-report requests rather than only technical-indicator analysis. In agent ecosystems, this can route unrelated or sensitive finance queries to an external service unexpectedly, increasing data exposure and misuse risk.
Natural-language policy review applies to all file types. The visible skill name and description are entirely in Chinese, and the usage examples are also Chinese-only, with no indication that users may choose another language or that the locale restriction is intentional and justified. This can conflict with organizational language-choice policies when skills are expected to avoid forcing a language without opt-in.