The Agent Skills Directory

[SAFE]: The skill is designed exclusively for structured documentation management and contains no executable code or dangerous command patterns.
[SAFE]: Data handling is restricted to the specific ARCHITECTURE.md file, with explicit instructions to avoid scanning the entire repository or guessing file paths.
[SAFE]: Security is enhanced by a mandatory human-in-the-loop requirement, where the AI is forbidden from automatically marking architectural decisions as confirmed ([x]) without explicit user sign-off.
[SAFE]: No external network dependencies, remote code downloads, or third-party package installations are present.

architecture-steward