The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines modules that ingest user-provided queries (e.g., question) and processes them through various DSPy modules.
Ingestion points: The forward methods in ComparisonPipeline, SequentialRAG, and RobustQA take a question string input.
Boundary markers: The skill uses DSPy Signature classes (BasicQA, QueryRewrite, GenerateAnswer, ValidateAnswer) which provide structural delimiters for inputs and outputs in the underlying LLM prompts.
Capability inventory: The skill is restricted to standard file operations (Read, Write, Glob, Grep) via the allowed-tools configuration in the YAML frontmatter.
Sanitization: No explicit sanitization is shown, though the use of structured signatures acts as a mitigating control by defining output expectations.
[EXTERNAL_DOWNLOADS]: The skill references the official DSPy library documentation and GitHub repository. These links point to well-known resources managed by the Stanford NLP Group and are provided for educational purposes.

dspy-advanced-module-composition