The Agent Skills Directory

[COMMAND_EXECUTION]: The file Openclaw一键安装.md contains a bash script designed to automate the installation of the necessary runtime environment. It executes commands to install nvm, Node.js, and the openclaw CLI, as well as managing local gateway services.
[EXTERNAL_DOWNLOADS]: The installation script downloads the nvm installer from raw.githubusercontent.com/nvm-sh/nvm/v0.40.0/install.sh. This is a well-known service for Node.js version management.
[REMOTE_CODE_EXECUTION]: The setup process involves piping a remote script from a well-known source (nvm) into bash and installing the openclaw package via npm.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from Xiaohongshu comments via references/xhs-comment-ops.md.
Ingestion points: Comments are retrieved from the /notification page of Xiaohongshu.
Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions when interpolating comment data into the agent's context.
Capability inventory: The skill has significant browser automation capabilities, including the ability to post notes and reply to comments using an authenticated profile.
Sanitization: There is no evidence of text sanitization or safety filtering for the content of the comments being processed.

xiaohongshu-ops