The Agent Skills Directory

[SAFE]: The primary logic in SKILL.md is strictly limited to structural scaffolding and configuration initialization, adhering to a 'no implementation code' policy that prevents the introduction of hidden malicious business logic.
[SAFE]: The core automation script scripts/scaffold.sh performs local file system operations using standard commands. It is designed to create directories and cat standard boilerplate content, with no external network dependencies or privilege escalation triggers.
[SAFE]: All external tools and libraries referenced (such as vite, create-next-app, and ruff) are established industry standards, and dependencies are sourced from trusted public registries.
[SAFE]: The skill implements protective measures against data leakage by automatically generating .gitignore files that ignore security-sensitive paths like .env, .aws, and .ssh credentials.
[SAFE]: Extensive reference documentation is provided for Staff+ engineers; while these files contain repetitive architectural filler text, they function as educational placeholders and contain no obfuscated malicious payloads or hidden instructions.

project-init