linkpix-clothing-recolor

Fail

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: HIGHPERSISTENCEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PERSISTENCE]: The skill provides instructions for the agent to append an export PATH statement to the user's ~/.bashrc file. This modification is intended to ensure that the manually installed Node.js binary is available in future shell sessions, but it constitutes a persistence mechanism.
  • [EXTERNAL_DOWNLOADS]: The skill downloads the Node.js binary and its SHA256 checksums from nodejs.org. It also includes fallback instructions to use npmmirror.com (an Alibaba-maintained mirror) if the official source is unreachable. These sources are well-known technology infrastructure providers.
  • [REMOTE_CODE_EXECUTION]: The skill instructions involve downloading a compressed archive (Node.js) and extracting it for execution. An automated scan flagged a command where a checksum file is piped to grep and sha256sum, which triggered a remote execution heuristic. While this specific command is a security integrity check, the overall process involves downloading and executing external binaries and scripts (e.g., via npm).
  • [COMMAND_EXECUTION]: The skill relies on the global installation and execution of the @iqinghu/qhkit npm package. It executes shell commands to perform image generation and configuration management (qhkit image generate, qhkit config set, etc.).
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to the following:
  • Ingestion points: Untrusted data enters the agent context via the uploadedImages (image files) and prompt (natural language text) parameters in the qhkit image generate command.
  • Boundary markers: There are no explicit delimiters or instructions provided to the underlying tools or models to ignore potentially malicious instructions embedded within the image metadata or the text prompts.
  • Capability inventory: The skill can perform network operations via the qhkit API and has file system access to read images and write configuration files.
  • Sanitization: No sanitization or validation of the input strings or image content is performed before interpolation into shell commands or processing by the generative AI model.
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 21, 2026, 04:09 PM