trace-to-training-data

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted production logs and evaluation traces to generate training data, which creates a surface for indirect prompt injection. If the source logs contain malicious instructions, they could be inadvertently incorporated into the training datasets.
  • Ingestion points: Untrusted data enters the context from runs/<run-id>/results.json and human-provided corrections.
  • Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the processed log entries.
  • Capability inventory: The skill performs JSON data transformations and filtering in Python. It does not invoke subprocesses or execute arbitrary commands based on the data content.
  • Sanitization: The skill explicitly requires a 'secret and PII scan' and redaction process (Hygiene section), but it does not specify filters for adversarial prompt injection content in the raw data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 07:33 PM
Security Audit — agent-trust-hub — trace-to-training-data