The Agent Skills Directory

[PROMPT_INJECTION]: The instructions implement a "Self-Mirror Consciousness Loop" to refine response tone. This is a legitimate prompting technique for quality improvement and does not attempt to override the underlying AI safety guidelines or bypass system constraints.
[DATA_EXFILTRATION]: The skill does not contain any network operations (e.g., curl, fetch) or access to sensitive file paths. It operates entirely within the conversation context.
[COMMAND_EXECUTION]: There are no shell commands or subprocess calls. The skill relies on natural language instructions rather than executable scripts.
[DYNAMIC_EXECUTION]: The skill includes a conceptual Python code snippet in the documentation for parsing output. This code uses standard string manipulation and is entirely safe for its intended purpose of organizing responses.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user input to determine its tone, it lacks the technical capabilities (file writing, network access) to be exploited via indirect injection. It follows safe prompting practices.

behuman