The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documents and provides examples for dspy.ProgramOfThought, a module designed to generate and execute Python code at runtime to solve problems. Additionally, a code example in references/examples.md demonstrates the use of eval() to process mathematical expressions, though it includes basic restrictions on the execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill extensively documents patterns for RAG (Retrieval-Augmented Generation) and agents that process external data from search tools or document databases. (1) Ingestion points: Untrusted data enters the agent context through dspy.Retrieve, search_wikipedia, and search_web functions as seen in references/examples.md and references/modules.md. (2) Boundary markers: Most examples do not demonstrate the use of explicit delimiters or instructions to ignore embedded commands in the retrieved content. (3) Capability inventory: The skill facilitates tool execution via dspy.ReAct and dynamic code execution via dspy.ProgramOfThought. (4) Sanitization: Examples lack explicit sanitization or validation of the data retrieved from external sources before it is interpolated into prompts.
[EXTERNAL_DOWNLOADS]: The installation instructions in SKILL.md include fetching the library directly from the Stanford NLP group's official GitHub repository (github.com/stanfordnlp/dspy.git).

dspy