learn-agent-skills

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches lesson manifests, documentation, and quizzes from a GitHub repository belonging to the author (fancyboi999/ai-engineering-from-scratch-zh). These are documented as educational materials and the domain (raw.githubusercontent.com) is a standard source for static assets.
  • [COMMAND_EXECUTION]: The skill instructs the agent to check for the presence of development tools such as node, npx, and python3 to ensure the learner's environment is ready for labs. It explicitly states that it will not install, update, or remove external systems without learner confirmation.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external lesson data (docs/zh.md and quiz.json). However, it contains specific instructions to treat skill and tool metadata as untrusted input (lesson 13/26) and requires explicit evidence/checkpoints, reducing the surface for accidental instruction obedience.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 05:59 AM
Security Audit — agent-trust-hub — learn-agent-skills