karpathy-system-prompt-learning

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists exclusively of educational content and templates for prompt engineering.
  • [REMOTE_CODE_EXECUTION]: No scripts or commands are present that would download or execute external code.
  • [DATA_EXFILTRATION]: No network tools or logic for data exfiltration were detected.
  • [CREDENTIALS_UNSAFE]: No sensitive file paths, secrets, or API keys are included in the instructions.
  • [PROMPT_INJECTION]: The content focuses on improving task performance through structured prompts and does not contain instructions to bypass AI safety guardrails or maliciously override agent behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 02:11 AM