The Agent Skills Directory

[SAFE]: The skill is a workflow manager for creating other skills within a controlled environment. It defines a structured process (Phase 1 to 6) for requirement gathering, research, and drafting.
[EXTERNAL_DOWNLOADS]: The skill references external research in 'Phase 1.5' using WebSearch and Context7 (an MCP tool). It provides a whitelist of trusted South Korean government and global regulatory domains (e.g., nts.go.kr, moel.go.kr, ifrs.org, eeoc.gov) for information gathering. These are standard information retrieval tasks for a research-oriented workflow and are considered safe.
[PROMPT_INJECTION]: No evidence of prompt injection or instructions to bypass safety filters. The skill actually includes safety-aligned instructions, such as a 'Korean Slop Prevention Rule' to ensure high-quality, human-like output in generated skills.
[COMMAND_EXECUTION]: The skill mentions external tools like WebSearch and Context7 (MCP), but these are used as intended for data retrieval within the agent platform's architecture. It does not attempt to execute arbitrary shell commands or perform unauthorized system modifications.
[DATA_EXFILTRATION]: There is no evidence of sensitive data access or transmission to unauthorized external servers. The workflow focuses on generating skill documentation and test cases based on user input and public research.

meta-skill-builder