The Agent Skills Directory

[Command Execution] (HIGH): The skill exposes $.child_process.exec in references/api-utilities.md, allowing the agent to execute arbitrary system commands. This is highly dangerous if model data is used to construct command strings without sanitization.
[Unverifiable Dependencies & Remote Code Execution] (MEDIUM): The load() function in references/api-utilities.md enables the dynamic execution of JavaScript files from the filesystem, which can be leveraged for RCE if file paths or content are influenced by untrusted sources.
[Data Exposure & Exfiltration] (MEDIUM): The skill provides unrestricted file writing via $.fs.writeFile in references/api-utilities.md, allowing for the potential exfiltration of sensitive model data to arbitrary local or network paths.
[Indirect Prompt Injection] (HIGH): The skill possesses a significant attack surface for indirect injection. Ingestion points: External ArchiMate models loaded via $.model.load (references/api-model.md). Boundary markers: Absent; the skill treats all model content as trusted data. Capability inventory: Powerful write and execute operations including $.child_process.exec and $.fs.writeFile. Sanitization: No validation or escaping is applied to untrusted model data before it is processed by high-privilege functions.

JArchi Scripting