The Agent Skills Directory

[COMMAND_EXECUTION]: The skill enables the agent to create and run arbitrary TypeScript code within the scripts/ directory using the pnpm script command. This is a core functionality that provides significant control over the execution environment.\n- [EXTERNAL_DOWNLOADS]: Examples provided in the documentation show how to use the fetch() API to retrieve content from external URLs, which is expected behavior for integrating with web services but involves remote data retrieval.\n- [PROMPT_INJECTION]: The skill introduces a surface for indirect prompt injection through its data processing templates. Scripts are designed to ingest untrusted data from files or network responses and report results back to the agent.\n
Ingestion points: Data enters the agent's context through fs.readFileSync in transform.ts and fetch() in generateImage.ts.\n
Boundary markers: The provided code examples do not use delimiters or instructions to prevent the agent from accidentally executing commands that might be embedded in the processed data.\n
Capability inventory: Scripts have access to the file system (read/write), network operations, and environment variables.\n
Sanitization: The templates lack any input validation, escaping, or sanitization logic for content parsed from external sources.

scripts