The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to utilize standard orchestration tools (Agent and Workflow) to delegate specific tasks to sub-models based on their complexity. For example, it routes code construction to Opus subagents and read-only chores to Sonnet.
[SAFE]: The skill implements a mandatory 'Review protocol' that requires the orchestrating model to read and verify the actual git diff produced by any subagent before reporting a task as complete. This serves as a critical human-in-the-loop or supervisor-level check on generated code.
[SAFE]: The instructions establish 'Escalation' rules where high-stakes tasks (authentication, payments, data integrity) are kept with the most capable model or require adversarial verification, demonstrating a security-conscious design.
[SAFE]: The skill uses a 'Context firewall' approach to prevent subagent logs or large artifacts from polluting the main session's context, which is a best practice for maintaining agent focus and resource efficiency.
[SAFE]: The installation methods provided in the README (npx and git clone) target the author's official GitHub repository and are standard practices for skill deployment.

fable-maestro