arize-prompt-optimization

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted production data (user inputs and LLM outputs) to generate optimized prompts, creating a surface for indirect prompt injection attacks.
  • Ingestion points: Production trace and experiment data exported to local JSON files (trace_/spans.json, dataset_/examples.json, experiment_*/runs.json).
  • Boundary markers: The meta-prompt template in references/optimization-meta-prompt.md uses delimiter headers like ORIGINAL BASELINE PROMPT and PERFORMANCE DATA to separate context from instructions.
  • Capability inventory: The skill uses the ax CLI for data export and experiment management, and jq for data transformation.
  • Sanitization: No explicit content filtering or sanitization of the performance data is mentioned before it is interpolated into the meta-prompt.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 09:07 AM
Security Audit — agent-trust-hub — arize-prompt-optimization