linkpix-1688-image

Fail

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: HIGHPERSISTENCEDYNAMIC_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PERSISTENCE]: The skill instructs the agent to persist the Node.js environment by appending export PATH commands to the user's shell configuration file (~/.bashrc).
  • [DYNAMIC_EXECUTION]: The skill provides a Python snippet (python -c) for image compression that uses string interpolation for file paths. This pattern is vulnerable to code injection if a file name contains single quotes or other control characters.
  • [DYNAMIC_EXECUTION]: The agent is instructed to use npx --yes sharp-cli to dynamically download and execute a Node.js package for image resizing at runtime.
  • [REMOTE_CODE_EXECUTION]: The environment setup process downloads Node.js binaries from official and mirror sources (nodejs.org, npmmirror.com) and extracts them for local execution. While the source is trusted, the automated execution of downloaded binaries is a significant capability.
  • [EXTERNAL_DOWNLOADS]: The skill fetches binaries, packages, and configuration files from several external sources including nodejs.org, the Alibaba npm mirror, and the Tsinghua University PyPI mirror. These are documented as well-known and trusted infrastructure providers.
  • [COMMAND_EXECUTION]: The skill executes multiple system commands to manage the environment, including curl for downloads, tar for extraction, sha256sum for integrity verification, and npm install for package management.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data via the customCopy and prompt parameters, which are then interpolated into commands sent to the qhkit tool and downstream AI models. It lacks explicit boundary markers to separate user instructions from system commands.
Recommendations
  • HIGH: Downloads and executes remote code from: https://nodejs.org/dist/v22.22.3/SHASUMS256.txt - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 11, 2026, 03:00 AM