Security audit
Security checks for vulnerabilities and agentic risk
This is a prompt-only workflow helper with some overly broad routing language, but no hidden code execution, persistence, credential access, or destructive behavior was found.
Install only if you want an advisory helper for Tavily or similar web-search workflow design. Expect possible accidental activation on generic web/search/API requests unless the publisher narrows the triggers, and avoid putting secrets or confidential data into any search-provider workflow you build from its guidance.
The skill description is so broad that it can match many ordinary requests involving work, web, search, APIs, checklists, or implementation help, causing unintended activation. Over-broad routing increases the chance that this skill intercepts unrelated tasks and applies the wrong workflow or constraints, which can degrade safety and reliability in agent systems.
The trigger list includes highly generic terms like 'web', 'search', 'api', 'user', and 'alternative', which are common across many benign conversations. In an agentic environment, these broad keywords can cause excessive false activations, leading to prompt hijacking of normal tasks, misrouting, and inappropriate application of this skill's instructions.
The trigger keyword list includes extremely broad, high-frequency terms such as "web", "search", "api", and "user", which can cause the skill to activate in many unrelated conversations. Over-broad activation increases the chance that this skill's instructions override or interfere with a more appropriate skill, producing unintended behavior and expanding the attack surface for prompt-routing abuse.
The trigger sentence begins with a broad phrase like 'Help me' and then appends a long requirement description, which creates a loose activation surface that may match many unrelated user requests. In an agent skill system, overly general triggers can cause incorrect skill invocation, routing user prompts into a web-search workflow when the user did not intend it, increasing the chance of unnecessary external-query behavior or irrelevant automation.
The phrase 'I need a practical workflow for ...' is a vague activation pattern because it can match many harmless productivity requests that have nothing to do with Tavily, web search, or this specific skill. In a multi-skill agent environment, such ambiguity can lead to mis-selection, causing the system to apply the wrong capability, potentially exposing user data to unnecessary processing paths or triggering external-search-oriented behavior unexpectedly.
The README promotes a web/search workflow 'via API' but does not warn that user queries, URLs, or related context may be transmitted to external services. This creates a data-handling transparency gap: users may provide sensitive internal information under the assumption it stays local, when the workflow could send it to third-party search providers.
The trigger phrases are broad enough to match ordinary requests about work productivity, web search, or bug fixing, which can cause the skill to activate outside its intended scope. In an agent ecosystem, overbroad routing can expose user prompts and workflow context to the wrong skill, increasing the chance of unintended tool use, external requests, or privacy-impacting behavior.
The trigger phrases are broad enough to match common, unrelated user requests, which can cause the skill to activate outside its intended scope. In an agent environment, overbroad activation can steer workflows toward unnecessary web/API usage, alter responses unexpectedly, and increase the chance that users are routed into a capability they did not explicitly request.
The example triggers are written as everyday language fragments and can match normal user requests that are not actually asking for this specific skill. Ambiguous examples train or encourage broad invocation behavior, which increases accidental activation and weakens routing precision.
The description says to use the skill when a user mentions broad concepts like work-productivity, web, search, or via, or whenever they need a practical workflow or analysis. This vague activation boundary makes routing unpredictable and can cause accidental invocation during ordinary requests that are not actually about Tavily-style workflow support.
The example trigger phrases begin with generic language like "Help me" and "I need a practical workflow," which closely matches normal user phrasing across many benign tasks. If used as routing exemplars, they can bias the system toward invoking this skill for unrelated prompts, leading to misfires and instruction collisions.
The default prompt is broad, generic, and composed of common-language phrases such as 'help me' and 'workflow helper,' which can cause the skill to activate in situations the user did not explicitly intend. In combination with a productivity/search-oriented skill, this raises the chance of accidental invocation, unexpected web-search behavior, and prompt-scope confusion across unrelated tasks.
Enabling implicit invocation without tight trigger boundaries allows the platform to invoke the skill based on loose semantic matches rather than explicit user selection. Because this skill relates to web/search workflows, unintended activation could expose user queries to unnecessary external-search handling, produce confusing behavior, or bypass user expectations about which agent capability is being used.
This file is labeled as a Chinese README, but key instructional content is presented in English, which can impose an unintended language preference on users of the zh-CN variant. There is no indication that users may choose their preferred language or that the mixed-language format is intentional and justified.
The description embeds the Chinese term '搜索-style' inside otherwise English activation text, which may imply a language-specific phrasing convention without explaining or offering language preference. While minor, the file does not state that users may choose their preferred language or locale.
The description and default prompt include English text mixed with the Chinese term "搜索-style" without indicating that the user can choose their preferred language or locale. This can conflict with language/locale policy expectations when a skill implicitly imposes multilingual output conventions without opt-in.
