building-agents-with-eve
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- External Resource Retrieval: The skill fetches lesson content and course metadata from
vercel.com. This is a well-known service and a trusted organization, which aligns with the skill's intended purpose as an Academy companion. - Codebase Progress Detection: To provide relevant guidance, the skill reads files within the student's
agent/andapp/directories. This allows the AI to determine which lesson the student is currently working on without requiring manual input. - Indirect Prompt Injection Surface: The skill is designed to follow instructions embedded within the fetched lesson markdown (using
<agent-instructions>tags). While processing external data for instructions is a known attack surface, the risk is mitigated by the use of a trusted source for the content. - Instructional Command Execution: The skill provides the student with terminal commands (e.g.,
npx,npm,curl) to set up and manage their local environment. These are presented as educational steps for the user to perform rather than the skill executing them autonomously.
Audit Metadata