The Agent Skills Directory

[SAFE]: The skill functions as a collaborative framework for designing and documenting new agent skills using natural language elicitation techniques.
[NO_CODE]: No executable code, binary files, or scripts are included within the skill; it operates exclusively through structured instructions and Markdown documentation.
[SAFE]: Explicitly forbids the generation of Python scripts or other executable code in the skills it creates, mitigating the risk of introducing remote code execution vectors in downstream artifacts.
[SAFE]: Implements a strict human-in-the-loop requirement for its self-improvement feature, ensuring that any substantive changes to instructions require explicit user confirmation before being applied.
[SAFE]: Features a mandatory 'Propagate Privilege Guidance' module that ensures legal confidentiality and attorney work product protections are addressed in every generated skill.
[SAFE]: References to platform-specific resources such as agentskills.io are documented neutrally and align with the skill's intended purpose for skill authoring and testing.

skill-creator