The Agent Skills Directory

[SAFE]: The skill implements a structured framework ('The Three Ms') designed to minimize risk by defaulting to the lowest possible autonomy level. It explicitly pushes back against fully autonomous agents (L4) and prioritizes deterministic (non-AI) code over AI-based solutions whenever possible.
[SAFE]: All generated artifacts include a mandatory 'bike-method-phase: 1' header, which requires users to manually validate and run the code before advancing to more automated phases. This serves as a safety guardrail against unintended automated actions.
[SAFE]: The skill includes an 'Intern Rule' in its framework, which advocates for best practices such as least-privilege access, scoped permissions, and never sharing personal credentials or passwords with the AI.
[SAFE]: File access is restricted to the specific context and configuration files defined in the skill's routes (e.g., priorities, connections, logs). It explicitly warns against creating duplicate context files and maintains existing content when performing repairs.
[SAFE]: While the skill can route to other tools like 'skill-creator' or 'skill-builder' to generate code, this is the intended primary purpose of the skill (assisting in workflow creation) and is triggered only after a multi-step human-in-the-loop scoping process.

level-up