andrej-karpathy-perspective
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill instructions mandate a specific role-play persona and define markers like 'STOP' and 'EXIT TRIGGER' to override default agent behavior and manage the role-play lifecycle.
- [COMMAND_EXECUTION]: The skill includes instructions for the agent to perform version checking by executing a shell command (
git ls-remote) to query tags from the author's repository. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon user-provided queries and external search results, creating a surface for indirect prompt injection. Ingestion points: Processes user questions and web search content; Boundary markers: Includes a disclaimer upon initial activation but lacks specific delimiters for search results; Capability inventory: The skill utilizes local command execution and web tools; Sanitization: No specific escaping or validation logic is defined for external content retrieved during the reasoning process.
- [EXTERNAL_DOWNLOADS]: The installation instructions recommend fetching the skill from the author's GitHub repository (
github.com/alchaincyf/karpathy-skill) and utilize a tool from a well-known service provider.
Audit Metadata