The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from existing skill and rule files which may contain embedded malicious instructions.
Ingestion points: The agent is instructed to use cat to read files from directories such as ~/.cursor/skills/, ~/.windsurf/rules/, and ~/.agents/skills/ into its context.
Boundary markers: Absent; there are no delimiters or instructions provided to the agent to prevent it from obeying instructions found within the managed files.
Capability inventory: The skill enables broad shell-based operations including file creation, modification, movement, and recursive deletion across numerous application configuration directories.
Sanitization: Absent; file content is ingested and processed without validation or filtering.
[COMMAND_EXECUTION]: The skill relies on direct shell command execution (ls, cat, mkdir, mv, cp, rm, find, grep) to manage files in the user's home directory. This provides the agent with significant autonomous control over the local file system.

manage-skills