Security audit
Security checks for vulnerabilities and agentic risk
The skill is mainly a browser automation guide, but it asks for broad execution authority and contains conflicting, overbroad instructions that go beyond local browser control.
Install only if you are comfortable giving an agent exec/read access and control of a browser session. Treat it as capable of acting on logged-in websites and possibly using a remote browser service if Browserbase environment variables are present; narrow or ignore the broad coding/deployment and file/API instructions before use.
The trigger condition is overly broad, stating the skill should be used for general coding, debugging, testing, and deployment scenarios well beyond browser automation. In an agent ecosystem, this increases the risk of unintended invocation, causing the skill to run exec/browser actions in contexts where users did not intend to authorize automation or command execution.
The manifest description uses broad productivity and workflow-optimization language without strong activation constraints. That wording can make the skill appear applicable to many unrelated tasks, increasing the chance that an agent selects it inappropriately and grants it access to browser or exec capabilities unnecessarily.
The file repeatedly states that the skill provides or supports Chinese interaction, including '支持中文交互' and related wording, but does not indicate that other languages are available based on user preference. This creates a locale/language constraint without clear user opt-in or justification as a region-specific tool.
The skill is presented as a local Chrome browser automation tool, but later claims broader capabilities such as file processing, API integration, and generic command execution. This capability drift can mislead an agent or user into granting or invoking actions outside the expected trust boundary, increasing the chance of unintended data access or command execution.
Declaring file read/parse/write style functionality in a browser automation skill expands the apparent authority of the skill beyond browser interaction. In an agent setting, this can cause the model to treat local file operations as in-scope and expose sensitive local data or modify files without users expecting that behavior.
Advertising external API integration in a skill primarily described as local Chrome automation widens the operational scope and trust assumptions. This can lead agents to send data to external services or request credentials in situations where users reasonably expected only local browser control.
The overview describes the free version as using local Chrome for rapid personal automation, implying a limited execution model. Later sections describe automatic switching to a remote browser mode with API keys and Browserbase configuration, which conflicts with the earlier characterization of the free edition's scope.
