ai-shifu-course-creator
Pass
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill transmits course content and metadata to the vendor's official domains, app.ai-shifu.cn and app.ai-shifu.com, via the scripts/shifu-cli.py utility. This communication is essential for the skill's primary function of deploying and managing live courses on the platform.
- [COMMAND_EXECUTION]: The agent is instructed to use a local CLI tool, scripts/shifu-cli.py, to perform all administrative tasks. This tool handles authentication, course building, and API interactions in a structured manner.
- [SAFE]: The skill demonstrates secure practices for managing sensitive information. It stores authentication tokens in a .env file and the CLI tool uses restricted file permissions (0o600) and path traversal checks to maintain local security.
- [PROMPT_INJECTION]: The skill processes untrusted external content (raw course materials) to generate scripts, which constitutes a vulnerability surface for indirect prompt injection.
- Ingestion points: Raw course material is ingested from user input or local files during the segmentation and orchestration phases.
- Boundary markers: The instructions do not currently specify the use of delimiters to wrap untrusted input material.
- Capability inventory: The skill can perform network operations and file writes through the provided CLI tool.
- Sanitization: There is no explicit requirement for sanitizing the input content before it is processed by the AI engine.
Audit Metadata