The Agent Skills Directory

[SAFE]: The skill implements a robust adversarial review process (Convention Warden) using a high-reasoning model (Opus) to validate all phase outputs, mitigating risks of architectural drift or insecure code generation.
[SAFE]: Explicit security principles are enforced (Shared Principle 4) that strictly prohibit agents from processing or outputting credentials, API keys, or PII, instructing them to flag secrets by location only.
[SAFE]: The skill maintains human oversight by requiring user validation for critical paths, test summaries, and architectural decisions, ensuring autonomous agents do not make sensitive system changes without approval.
[SAFE]: Strict write-safety and role-scoping rules are established to prevent concurrent agents from interfering with shared project files or overwriting each other's progress logs.
[SAFE]: Tech stack detection and file system operations are limited to standard project reconnaissance and directory scaffolding required for the skill's stated engineering purpose.

craft-laravel