linkpix-image-eraser

Fail

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: HIGHPERSISTENCEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PERSISTENCE]: The skill provides instructions for the agent to modify the user's shell configuration file (~/.bashrc) to persist changes to the PATH environment variable.
  • Evidence: export PATH="$HOME/.local/lib/node-v22.22.3-linux-x64/bin:$PATH" (或追加进 ~/.bashrc) in SKILL.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it interpolates untrusted user instructions directly into a shell command for the qhkit tool.
  • Ingestion points: User-provided descriptions of elements to be erased (e.g., "watermark", "text") are inserted into the prompt field of the command-line arguments in SKILL.md.
  • Boundary markers: The JSON payload is wrapped in single quotes, but there is no explicit instruction to sanitize or escape user-provided strings.
  • Capability inventory: The qhkit CLI tool has capabilities for network communication with vendor APIs and local file system access for image uploads.
  • Sanitization: The skill lacks any sanitization or validation logic for the interpolated user strings.
  • [EXTERNAL_DOWNLOADS]: The skill automates the download of binaries and software packages from external sources to bootstrap its environment.
  • Fetches Node.js binaries from nodejs.org (a well-known service).
  • Downloads the @iqinghu/qhkit package from the NPM registry or the npmmirror.com mirror.
  • Note: The skill uses sha256sum to verify the integrity of the Node.js binary, which is a defensive best practice.
  • [COMMAND_EXECUTION]: The skill relies on various shell commands to perform setup and core functions, including npm, curl, tar, and the vendor-specific qhkit CLI tool.
  • These commands are used to manage environment configuration and execute image processing tasks.
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