The Agent Skills Directory

EXTERNAL_DOWNLOADS (HIGH): The skill uses npx to download and execute the @superdoc-dev/cli package from the npm registry. This author and organization are not part of the trusted sources list, making the execution of this code a significant security risk.
REMOTE_CODE_EXECUTION (HIGH): By specifying @latest in the npx commands, the skill ensures that the most recent version of the package is fetched and run every time. This is a dangerous practice as an attacker compromising the @superdoc-dev npm account could push a malicious update that would be automatically executed by any agent using this skill.
COMMAND_EXECUTION (MEDIUM): The skill's primary function is to execute shell commands. While intended for document processing, the underlying tool has full access to the shell environment where it is executed.
PROMPT_INJECTION (LOW): As the skill is designed to 'read' and 'search' Word documents, it is vulnerable to Indirect Prompt Injection (Category 8). If a processed .docx file contains hidden instructions, the agent might inadvertently follow them after the text is extracted. No boundary markers or sanitization logic is present to mitigate this.

docx-processing-superdoc