qinghu-detail-page-clone

Fail

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: HIGHPERSISTENCEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PERSISTENCE]: The skill includes instructions to modify the ~/.bashrc shell profile to persistently add a custom Node.js installation directory to the system PATH environment variable.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content that is interpolated into prompts for image generation.\n
  • Ingestion points: User-provided selling point text (卖点文案) and uploaded product/reference images (uploadedImages) in SKILL.md.\n
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when incorporating user-supplied text into the model prompt.\n
  • Capability inventory: The skill executes the qhkit CLI tool, which performs network operations to external APIs and triggers image generation.\n
  • Sanitization: There is no evidence of sanitization or validation of the user-provided text before it is interpolated into shell commands and prompts.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to bootstrap the environment, including npm install, tar, and mkdir. It also constructs shell commands by interpolating user-supplied text into a JSON string for the qhkit tool, which presents a command injection risk if user input contains shell-sensitive characters like single quotes.\n- [EXTERNAL_DOWNLOADS]: The skill downloads the Node.js runtime from nodejs.org and the @iqinghu/qhkit package from the NPM registry. These downloads originate from well-known services and the vendor's official package repository.
Recommendations
  • HIGH: Downloads and executes remote code from: https://nodejs.org/dist/v22.22.3/SHASUMS256.txt - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 11, 2026, 03:00 AM