The Agent Skills Directory

[SAFE]: The skill defines behavioral templates for the AI agent to optimize output for different task types (e.g., Debugging, Code Review, Implementation). These templates guide communication style and prioritization without introducing executable logic or dangerous commands.
[SAFE]: Tool access is explicitly restricted in the YAML frontmatter to basic file reading operations (Read, Glob, Grep), preventing unauthorized file modification or network communication.
[SAFE]: The "SHIP" mode includes security-positive instructions, requiring the agent to verify the absence of exposed secrets and ensuring input validation is complete before considering a task finished.
[SAFE]: All instructions are transparent, using standard markdown for formatting guidelines without any hidden code, obfuscation, or persistence mechanisms.

behavioral-modes