jfrog-init
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on executing local Node.js scripts and system commands (node, npx, uname, winget, open/start) to detect the environment state and perform setup tasks.
- [EXTERNAL_DOWNLOADS]: Fetches the JFrog CLI binary from the vendor's own domain (releases.jfrog.io) and the Node Version Manager (nvm) installation script from a well-known community repository on GitHub (nvm-sh/nvm). These are trusted and expected sources for the skill's functionality.
- [REMOTE_CODE_EXECUTION]: On macOS and Linux, the skill provides instructions to install Node.js by piping the nvm installer script directly from GitHub into a shell. This is an industry-standard installation method for the intended task.
- [PERSISTENCE]: Modifies shell configuration files (e.g., .bashrc, .zshrc, config.fish) to persist the JFrog CLI's location on the system PATH. This behavior is clearly disclosed to the user during the interactive installation process.
- [INDIRECT_PROMPT_INJECTION]: Processes user-supplied project names and keys which are used to construct API URLs and update configuration files. The skill implements robust regex-based sanitization (e.g., SAFE_URL, SERVER_ID, SHELL_SYNTAX) to prevent command or URL injection attacks.
- [DYNAMIC_EXECUTION]: Generates a local helper script (marketplace-headers.mjs) and configures it as a headersHelper within Claude Code's settings. This script is executed dynamically by the host to retrieve authentication headers from the JFrog CLI, ensuring tokens are handled in memory and not stored in plain text configuration files.
Audit Metadata