The Agent Skills Directory

[Remote Code Execution] (CRITICAL): The skill requires the mh CLI and provides an installation command that downloads a shell script from an untrusted GitHub repository (vaayne/mcphub) and pipes it directly into the shell (curl -fsSL ... | sh). This allows for immediate, arbitrary code execution on the user's machine from an unverified source.
[Command Execution] (HIGH): The tool executeTerminalCommand allows the agent to execute any shell command within the IDE terminal. Combined with executeRunConfiguration, this provides a direct path for system compromise if the agent is misled by malicious instructions.
[Indirect Prompt Injection] (HIGH): The skill has a massive attack surface for indirect prompt injection.
Ingestion points: getFileTextByPath, searchInFilesByText, and listDirectoryTree ingest untrusted content from the project files.
Capability inventory: Includes arbitrary command execution (executeTerminalCommand), file creation (createNewFile), and file modification (replaceTextInFile).
Sanitization/Boundaries: No sanitization or boundary markers are defined.
Risk: An attacker could place a malicious instruction inside a source code file (e.g., in a comment). When the agent reads that file for context, it may execute the embedded command via the IDE's terminal.
[Data Exposure] (HIGH): Tools like getFileTextByPath and getAllOpenFilePaths allow the agent to read any file within the project scope. If the IDE is misconfigured or has broad permissions, this could lead to the exposure of sensitive files like .env files, SSH keys, or cloud credentials.

mcp-jetbrains-ide