The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): The content is purely informational documentation and contains no instructions to bypass safety protocols or override agent system prompts.
[DATA_EXFILTRATION] (SAFE): There are no hardcoded credentials, sensitive file path targets (such as .ssh or .aws), or logic for exfiltrating local data.
[EXTERNAL_DOWNLOADS] (SAFE): Examples of network tools like curl use generic domains and do not include instructions for automated remote code execution or untrusted script sourcing.
[COMMAND_EXECUTION] (SAFE): The bash patterns provided are standard administrative commands for file management, process handling, and text processing, with no evidence of destructive or obfuscated payloads.

bash-linux