The Agent Skills Directory

[SAFE]: The skill is designed for requirement gathering and documentation within a workspace. Its instructions focus on structured interviewing ('grilling') to define workflow specifications.
[EXTERNAL_DOWNLOADS]: The skill documentation (UPSTREAM.md) contains links to the original source code and maintenance repositories on GitHub. These references are to a well-known developer and service, and are used for tracking skill provenance rather than unauthorized data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from NOTES.md and workflows/*.md. While these files could theoretically contain instructions, the skill's primary function is to ask questions and generate documentation. The absence of high-privilege execution tools in the provided scripts minimizes the risk associated with this ingestion surface.
[DATA_EXPOSURE]: The skill interacts with workspace files to store and retrieve user-defined workflow specs. This behavior is localized and consistent with the skill's stated purpose as a design tool.

loop-me