The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No instructions designed to override agent behavior or bypass safety guardrails were found. The skill contains standard technical documentation and Python code snippets.
[DATA_EXFILTRATION] (SAFE): The skill interacts with standard LLM endpoints and a specific public research index (ColBERTv2) at 20.102.90.50, which is standard for the DSPy ecosystem. No access to sensitive user files or credentials was observed.
[EXTERNAL_DOWNLOADS] (SAFE): The skill does not download external scripts or execute remote code at runtime. It relies on the 'dspy' package, which is a legitimate library for AI program optimization.
[COMMAND_EXECUTION] (SAFE): No shell commands, subprocess spawning, or arbitrary code execution patterns are present. The skill strictly uses Python APIs for its intended purpose.
[INDIRECT_PROMPT_INJECTION] (LOW):
Ingestion points: The skill processes training datasets (trainset) and external search results from a ColBERTv2 server.
Boundary markers: Delimiters are not explicitly defined in the example prompt templates.
Capability inventory: The skill is granted 'Read' and 'Write' tools in its configuration, and the code performs network retrieval and local file writing for model weights and metadata.
Sanitization: No specific sanitization of the training data or search results is demonstrated in the examples.
[FALSE_POSITIVE_ANALYSIS] (SAFE): The 'Malicious URL' alert for 'logger.info' is a false positive generated by the scanner's failure to distinguish between a Python method call and a URL structure. 'logger.info' is a standard logging command and poses no security risk.

dspy-miprov2-optimizer