The Agent Skills Directory

[SAFE]: The skill's functionality is limited to conversational interaction and managing markdown files in the workflows/ directory. It contains no instructions for network access, remote script execution, or privilege escalation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from user notes (NOTES.md) and conversation to generate specifications. While this is an ingestion surface for indirect instructions, the skill lacks dangerous capabilities such as shell execution or network exfiltration, rendering the risk negligible and standard for its intended purpose.

loop-me