dspy-optimize-anything

Warn

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions specify the installation of the 'gepa' Python package via pip, which is a dependency from a source outside of the pre-defined trusted vendors.
  • [COMMAND_EXECUTION]: In the 'evaluate_solver' example, the skill uses 'subprocess.run' to execute Python code strings generated by the AI optimizer. This creates a surface for arbitrary system command execution if the generation process is influenced by malicious input.
  • [REMOTE_CODE_EXECUTION]: The 'evaluate_agent' example utilizes the 'exec()' function to dynamically run generated Python functions. This allows for arbitrary code execution within the agent's runtime environment.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via the 'objective' and 'background' fields. * Ingestion points: Data enters through the 'objective' and 'background' parameters defined in the optimization call. * Boundary markers: There are no boundary markers or instructions to ignore embedded commands within the processed text artifacts. * Capability inventory: The skill possesses 'Read', 'Write', and 'Grep' tools, alongside 'subprocess' and 'exec' capabilities. * Sanitization: No input validation or sanitization is performed on the natural language instructions before they are used to guide the code generation proposer.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 25, 2026, 03:11 AM