karpathy-education-first

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill is entirely composed of markdown-formatted prompts and pedagogical instructions designed to help an AI agent explain technical concepts or simplify code.
  • [NO_CODE]: No scripts, shell commands, or binary executables are included in the skill. It relies purely on natural language instruction.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides templates that ingest external data (e.g., [PASTE CODE], [PASTE DOCUMENTATION], [PASTE ABSTRACT]).
  • Ingestion points: SKILL.md contains several prompt templates for processing user-supplied code, research, and documentation.
  • Boundary markers: Absent; the templates use simple placeholders without explicit delimiters or instructions to ignore embedded commands.
  • Capability inventory: No sensitive tools, file system access, or network capabilities are requested in the skill frontmatter or body.
  • Sanitization: None detected.
  • Note: While an attack surface for indirect prompt injection exists, the lack of actionable capabilities (tools) renders this risk very low.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 02:11 AM