rlm-gepa

Pass

Audited by Gen Agent Trust Hub on Aug 5, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill provides a legitimate development framework for optimizing AI agent instructions. It includes documentation on best practices for data management and evaluation safety.\n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the predict-rlm package. This is a vendor-provided dependency required for the language model optimization functionality.\n- [PROMPT_INJECTION]: The skill defines a process for optimizing instructions using external datasets, which introduces a surface for indirect prompt injection.\n
  • Ingestion points: Data is loaded via load_trainset() and load_valset() in the project implementation.\n
  • Boundary markers: The skill explicitly guides users to audit data, use deterministic splits, and prevent data leakage.\n
  • Capability inventory: The optimized agent executes logic using the PredictRLM class and declared tools.\n
  • Sanitization: The workflow relies on user-led data auditing to ensure the safety of the training and validation sets.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 5, 2026, 11:44 PM
Security Audit — agent-trust-hub — rlm-gepa