qiaomu-opencli-explorer
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documents and uses commands like
opencli generateandopencli synthesizewhich automatically create TypeScript code based on the analysis of external websites. It further describes a mechanism where TypeScript adapters placed in specific directories are "automatically dynamically registered" and executed by the tool at runtime. - [DYNAMIC_EXECUTION]: Extensive instructions are provided for using
page.evaluateto execute arbitrary JavaScript within the browser context. This includes complex logic for interceptingfetchandXMLHttpRequestcalls, as well as interacting directly with frontend framework states like Pinia stores to trigger actions and retrieve data. - [COMMAND_EXECUTION]: The skill defines a workflow involving several CLI commands (
opencli explore,opencli record,opencli cascade) that perform automated web navigation, network monitoring, and system-level interactions to discover API endpoints. - [PROMPT_INJECTION]: The guide includes explicit directives to the AI agent (e.g., "You (AI Agent) must through browser open target website to explore!", "AI Agent exploration workflow (must follow)") designed to override standard agent behavior and enforce a specific operating procedure for web exploration.
- [INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection as it is designed to scrape and process data from arbitrary, untrusted websites. The instructions do not provide a framework for sanitizing or isolating this external content before it is returned to the agent's context.
- Ingestion points: Arbitrary target URLs processed via
browser_navigateandopencli explore. - Boundary markers: None identified in the workflow to separate external site content from agent instructions.
- Capability inventory: The tool possesses capabilities for network requests, file system writes (adapter creation), and shell command execution.
- Sanitization: General sanitization instructions are absent, with only a basic
stripHtmlutility mentioned for specific platform use cases. - [EXTERNAL_DOWNLOADS]: The skill encourages the installation and use of external software from the
jackwener/openclirepository and associated Node.js packages.
Audit Metadata