The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the myagents CLI to manage document conversion jobs. These commands (convert, status, wait, cancel, list) are standard for local task management and are scoped to the skill's primary purpose.
[DATA_EXPOSURE_AND_EXFILTRATION]: The instructions explicitly include a security boundary regarding credential safety, directing the agent not to repeat or log document passwords provided by the user. It emphasizes offline operation and limits input to local files, specifically forbidding the use of URLs or remote sources.
[INDIRECT_PROMPT_INJECTION]: As a document parsing utility, the skill inherently processes untrusted data from local Office, PDF, and image files. While this represents a potential attack surface for indirect prompt injection, the skill defines clear boundaries for the agent and relies on a specific local CLI tool for processing rather than passing raw file content directly into the model prompt.
[REMOTE_CODE_EXECUTION]: No remote script downloads or installations were detected. The skill explicitly states that it does not require system-level Python/Node, Docker, or runtime internet connectivity, relying instead on the pre-installed local application.

myagents-anydoc