The Agent Skills Directory

[COMMAND_EXECUTION]: The start_node_debug tool allows the agent to spawn new Node.js processes on the host system. It accepts absolute script paths, command-line arguments, and environment variables, providing a direct interface for process execution.
[DYNAMIC_EXECUTION]: The evaluate_expression tool enables the execution of arbitrary JavaScript code within the context of a paused Node.js process. This allows the agent to dynamically interact with and modify the runtime state of the application being debugged.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by processing external and potentially untrusted data that could contain malicious instructions designed to influence the agent.
Ingestion points: Application source code retrieved via get_script_source, console logs captured by read_console, and runtime variable data accessed through inspect_scopes and get_object_properties.
Boundary markers: The skill instructions do not specify the use of delimiters or clear separation between system instructions and data retrieved from the debugged process.
Capability inventory: The agent has the ability to launch processes (start_node_debug), read local files (get_script_source), and execute arbitrary code within the target process (evaluate_expression).
Sanitization: There is no evidence of filtering or sanitizing the content retrieved from the Chrome DevTools Protocol before it is presented to the agent.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions to install the devtools-debugger-mcp package from the NPM registry, which is a standard method for distributing Node.js tools.

devtools-debugger-mcp-nodejs