The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and installs code from unverified third-party sources.
Clones a repository from https://github.com/jkf87/hwp-mcp.git.
Installs the NPM package @ohah/hwpjs globally.
[REMOTE_CODE_EXECUTION]: The skill executes installation routines that run code from untrusted external repositories.
Runs pip install -r requirements.txt immediately after cloning an untrusted repository (hwp-mcp). This can execute arbitrary code during the installation process via malicious requirements or dependency hooks.
Installs and executes a global Node.js package (@ohah/hwpjs) from an unverified publisher.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its core functionality and setup.
Executes npm install -g, git clone, and pip install to prepare the environment.
Uses node -p to probe system information.
Invokes hwpjs and Python-based servers to perform document conversions and automation.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8) as it processes untrusted HWP documents while possessing high-privilege capabilities.
Ingestion points: Processes .hwp files from user-provided paths (SKILL.md).
Boundary markers: None present; the skill does not use delimiters or instructions to ignore embedded commands within the processed files.
Capability inventory: Includes package installation (npm, pip), repository cloning (git), and file system write operations.
Sanitization: No sanitization or validation is performed on the content extracted from HWP files before it enters the agent context.

hwp