The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill documents the DSPy::CodeAct module in references/core-concepts.md. This feature allows the LLM to generate and execute Ruby code at runtime to solve tasks. While intended for problem-solving, executing LLM-generated code is a high-risk activity that can lead to arbitrary code execution if the environment is not properly sandboxed.
[PROMPT_INJECTION] (LOW): The skill demonstrates processing untrusted data through templates in assets/module-template.rb. This creates a surface for indirect prompt injection.
Ingestion points: email_subject, email_body, and task parameters in assets/module-template.rb and assets/signature-template.rb.
Boundary markers: No explicit boundary markers (e.g., XML tags or delimiters) are implemented in the templates to isolate user input from instructions.
Capability inventory: The AgentModule utilizes several tools (SearchTool, CalculatorTool, DatabaseQueryTool) which could be subject to manipulation through injected commands in the input data.
Sanitization: No input sanitization or validation logic is present in the templates to filter potentially malicious instructions.
[EXTERNAL_DOWNLOADS] (LOW): The references/providers.md file recommends the installation of various dspy-* gems. While these are legitimate library extensions, users should ensure they are sourced from trusted registries. This is noted as low risk as it follows standard library dependency patterns.

dspy-ruby