human-in-the-loop-training

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a system architecture designed to ingest untrusted human feedback and process it through LLMs to generate new rules or training data.
  • Ingestion points: The FeedbackCollector, HumanReviewQueue, and PatternLearner classes in SKILL.md are designed to handle external human input.
  • Boundary markers: The provided code snippets lack explicit boundary markers or instructions to the LLM to ignore embedded commands within the human corrections.
  • Capability inventory: The design suggests capabilities for triggering model training jobs (submit_training_job) and generating text based on corrections (llm.generate).
  • Sanitization: There is no evidence of sanitization or filtering for the human-provided feedback before it is interpolated into prompts in the derive_rule and generate_with_corrections methods. This represents a vulnerability surface where a malicious reviewer could inject instructions that influence model behavior or rule generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:04 PM
Security Audit — agent-trust-hub — human-in-the-loop-training