The Agent Skills Directory

[SAFE]: The skill is focused on documentation management, specifically creating and maintaining Architecture Decision Records in markdown format.\n- [SAFE]: File operations are limited to the repository's design-docs/ directory and the CLAUDE.md configuration file for the purpose of maintaining a documentation index.\n- [SAFE]: No instances of prompt injection, obfuscation, or data exfiltration attempts were detected.\n- [SAFE]: The skill does not download external dependencies or execute remote code; it utilizes local assets provided within the skill directory.\n- [SAFE]: Explicit safety instructions are included, such as requiring the agent to 'Stop and ask when a task conflicts with an ADR', which ensures human-in-the-loop oversight for significant changes.

adr