The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Requires global installation of the unverified @jackwener/opencli npm package and a browser extension from a third-party GitHub repository.\n- [REMOTE_CODE_EXECUTION]: The opencli plugin install command allows the agent to download and execute code from arbitrary GitHub repositories or npm packages.\n- [COMMAND_EXECUTION]: The opencli external register feature enables the agent to wrap and execute sensitive system binaries (e.g., gh, docker, kubectl), increasing the attack surface if the tool is misused.\n- [DYNAMIC_EXECUTION]: The opencli browser eval primitive allows for the execution of arbitrary JavaScript code within a browser session.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the web via extract, find, and network primitives, creating a surface for indirect prompt injection.\n
Ingestion points: Data extracted from web pages via browser extract and network monitoring.\n
Boundary markers: The skill documentation does not provide instructions for the agent to use delimiters or ignore instructions found in web data.\n
Capability inventory: Includes execution of JavaScript (eval), interaction with browser sessions (click, type), and execution of local binaries (external register).\n
Sanitization: No explicit sanitization or filtering of web content is defined before processing.

opencli-universal-cli-hub