The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/download.sh uses the eval command to execute a string constructed from user-supplied arguments, specifically the URL, token, and key values. This allows for arbitrary command execution if an attacker provides inputs containing shell metacharacters such as backticks, semicolons, or dollar-sign parentheses (e.g., providing a token like $(whoami)).
[CREDENTIALS_UNSAFE]: The skill's instructions and script encourage users to manually extract and provide their _yuque_session cookie. While necessary for the tool's functionality, this credential provides full access to the user's Yuque account and is handled unsafely within the command-line arguments.
[COMMAND_EXECUTION]: The skill relies on the global installation of a third-party package yuque-dl via npm. The agent is instructed to execute this tool directly, which increases the attack surface if the package or the environment is compromised.
[COMMAND_EXECUTION]: The SKILL.md and scripts/download.sh files contain a hardcoded absolute path /Users/lytton/mac_wps_clound/Obsidian笔记知识库/宝藏资源库/语雀知识库/. This path exposes the local username 'lytton' and the user's private directory structure, which is a data exposure concern.

yuque-dl