dspy-debugging-observability
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references a public ColBERTv2 retrieval endpoint hosted on Azure (http://20.102.90.50:2017/wiki17_abstracts) for data retrieval examples. This is a standard demonstration service frequently used in the DSPy community for research and educational purposes.
- [INDIRECT_PROMPT_INJECTION]: The skill documents a Retrieval-Augmented Generation (RAG) architecture, which involves ingesting data from external services that could theoretically contain malicious instructions.
- Ingestion points: External passages are fetched via the
dspy.ColBERTv2client in the RAGPipeline example. - Boundary markers: The examples focus on observability and tracing; they do not explicitly demonstrate prompt-level delimiters or sanitization for fetched content.
- Capability inventory: The skill configuration allows for basic file operations via
Read,Write,Glob, andGreptools. - Sanitization: The skill includes a best practice recommendation to redact Personally Identifiable Information (PII) before logging traces to external monitoring systems.
Audit Metadata