The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill employs standard piped execution patterns (e.g., curl | sh and bash -c "$(curl...)") to install packages. The sources include official project domains such as astral.sh for the uv tool and raw.githubusercontent.com/Homebrew for the Homebrew package manager. It also provides academic mirrors from the University of Science and Technology of China (mirrors.ustc.edu.cn) for users in specific network environments. These are recognized, legitimate installation vectors for these tools.
[EXTERNAL_DOWNLOADS]: The skill downloads installers and binaries from official vendors including Node.js, Python.org, and Go.dev. It also references GitHub-hosted projects for various CLI utilities. These downloads are performed to fulfill the skill's primary purpose of environment setup.
[COMMAND_EXECUTION]: Extensive use of shell and PowerShell commands to perform system discovery (e.g., uname, sw_vers), check for existing tool versions, and execute package managers. This behavior is necessary for diagnostic and installation tasks.
[PRIVILEGE_ESCALATION]: Uses sudo on macOS for operations that require elevated permissions, such as using the system installer for .pkg files or softwareupdate for Xcode Command Line Tools. It includes logic to detect if sudo is available without a password before attempting non-interactive execution.
[PERSISTENCE]: Modifies shell profile files (~/.zshrc, ~/.zprofile) and system environment variables (Windows PATH) to ensure installed tools remain accessible across sessions. These modifications are standard practice for software installation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external command outputs, such as version numbers and diagnostic messages. While this provides a potential surface for indirect injection, the risk is considered low as the data is used for version comparison and user feedback within a controlled workflow.

qclaw-env