The Agent Skills Directory

[SAFE]: The skill's primary purpose is to assist users in authoring SKILL.md files. It includes robust instructions for a 'Guided Wizard', 'Conversation-to-Skill' extraction, and 'Review/Improve' modes. All logic is focused on prompt engineering and workflow automation.
[EXTERNAL_DOWNLOADS]: The skill references the official Anthropic skills repository (github.com/anthropics/skills) for documentation and starter templates. These are trusted sources and integral to the skill's function.
[COMMAND_EXECUTION]: The documentation mentions using scripts for deterministic tasks (e.g., math, file parsing). While it provides a framework for execution, the skill itself does not ship with malicious scripts or unauthorized command patterns.
[PROMPT_INJECTION]: Analysis of the instructions confirms they do not attempt to bypass LLM safety guardrails. In fact, the skill includes a 'Principle of Lack of Surprise' section explicitly forbidding malware or exploit code.
[DATA_EXFILTRATION]: No network operations to unknown domains or sensitive file access patterns were detected. The skill focuses on processing user-provided conversation history or manual inputs.

create-skill