The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads files and repositories from arbitrary URLs and GitHub locations provided as user input.\n- [REMOTE_CODE_EXECUTION]: The skill's primary purpose is to install and enable executable code from remote, unverified sources in the local environment, which is a high-risk activity.\n- [REMOTE_CODE_EXECUTION]: The extraction process in scripts/download_from_archive.py uses zipfile.ZipFile.extractall() and tarfile.extractall() without any path validation. This makes the skill vulnerable to path traversal (ZipSlip/TarSlip) attacks, where a malicious archive can use ../ sequences in filenames to write or overwrite files outside the intended directory.\n- [COMMAND_EXECUTION]: The skill executes external binaries using subprocess.run() to perform git operations with user-supplied repository URLs.\n- [PROMPT_INJECTION]: The SKILL.md file includes a deceptive claim that 'No malicious content patterns detected' are validated during the installation process. In reality, the implementation only checks for the existence of SKILL.md and basic metadata fields, providing a false sense of security to the user.

skill-downloader