The Agent Skills Directory

[DYNAMIC_EXECUTION]: The module facilitates the generation and execution of Python code by an LLM to explore large contexts. The documentation specifies that the execution occurs within a sandboxed Pyodide/WASM environment managed by Deno, which lacks network and filesystem access by default.
[COMMAND_EXECUTION]: The reference.md file provides examples of how users can extend the module's capabilities using custom tools, such as a file reader or regex search. The skill includes a clear 'Security note' warning that the security posture of these tools is the user's responsibility and advises against exposing dangerous functions like subprocess.run.
[INDIRECT_PROMPT_INJECTION]: The module is designed to ingest large, external contexts such as logs or codebases. The documentation provides a best-practice recommendation to pass this data as keyword arguments to the module rather than concatenating it into the prompt, which helps isolate untrusted data from the agent's instructions.
[EXTERNAL_DOWNLOADS]: The skill identifies Deno as a prerequisite for the sandbox environment and provides links to its official website. Deno is a well-known, security-focused runtime, and its inclusion is standard for the functionality described.

dspy-rlm-module