The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs software from sources that are not categorized as trusted. It performs a global npm installation of @ohah/hwpjs and clones a Git repository from https://github.com/jkf87/hwp-mcp.git.
[REMOTE_CODE_EXECUTION]: The workflow involves cloning an external repository and then executing its contents (e.g., hwp_mcp_stdio_server.py) after installing dependencies via pip install -r requirements.txt. This represents the execution of arbitrary code from an external provider.
[COMMAND_EXECUTION]: The skill uses the shell to detect the operating system platform (node -p "process.platform") and execute various CLI tools (hwpjs) with arguments derived from input files.
[INDIRECT_PROMPT_INJECTION]: The skill represents an indirect injection surface because it processes untrusted .hwp documents while possessing powerful system capabilities.
Ingestion points: Reads .hwp files from local paths for conversion and processing (SKILL.md).
Boundary markers: None present to delimit untrusted file content or prevent instruction following from within documents.
Capability inventory: Performs npm install, git clone, pip install, and executes shell commands for document conversion and image extraction (SKILL.md).
Sanitization: No sanitization or validation of the .hwp content is specified before processing.

hwp