huashu-design
Fail
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides instructions to install another skill from an external, unverified repository (alchaincyf/huashu-design) using the 'npx skills add' command. This pattern facilitates the execution of remote code that has not been audited.
- [COMMAND_EXECUTION]: The decision tree section includes shell scripts and commands for the agent to execute, such as checking for file existence, using 'grep', and performing installation routines.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill processes design prototypes, user-selected design directions, and maintains a 'look ledger' of shipped projects in SKILL.md.
- Boundary markers: No specific delimiters or safety instructions are provided to isolate processed design data from the agent's control flow.
- Capability inventory: The skill utilizes helper scripts (learn.py, design_lint.py, capabilities.py) and various Node.js libraries for file generation and export (Playwright, ffmpeg) as listed in the requirements.
- Sanitization: There is no evidence of input validation or sanitization for the data being processed by the helper scripts.
Recommendations
- AI detected serious security threats
Audit Metadata