course-guide
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches course metadata, README files, and glossary terms from the author's GitHub repository (raw.githubusercontent.com/fancyboi999/ai-engineering-from-scratch-zh) when local versions are unavailable. These resources are hosted on a well-known service and belong to the skill author's infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from external documentation to determine routing logic and agent behavior, creating a potential attack surface if those files are compromised.
- Ingestion points: Reads README.md, docs/zh.md, and JSON manifest files (e.g., model-context-protocol.json, program.json) from the local environment or remote repository as specified in the routing instructions.
- Boundary markers: None identified; the agent is instructed to use the content directly for routing and recommendation decisions without specific delimiters for external content.
- Capability inventory: Summarizes content and recommends executable commands (such as 'learn', 'check-understanding', and 'start-learning') to the user based on processed text.
- Sanitization: No explicit sanitization or validation of the ingested documentation content is performed before processing or interpolation into agent responses.
Audit Metadata