Interview me skill

The video demonstrates the use of the "Interview Me" skill within Cloud Code, applied to a real-world project called Loadout (formerly Toolbox), a desktop application for managing coding skills and configurations. The goal is to add project-specific skill management, as currently Loadout only supports global installations.

The user initiates an "Interview Me" session, focusing on the product aspect first. The AI asks clarifying questions about the target user, pain points, and desired functionality to understand the problem space. The AI helps to refine the user's thinking by prompting consideration of various design choices, such as how to handle globally installed skills in the context of a specific project.

Key decisions and insights that emerge include:

• Skills are cached internally but deployed through symlinks, either globally or project-specific.
• Users will manually add project directories to Loadout.
• The UI will have a scope picker (likely a combo box) defaulting to "global", allowing users to switch to specific projects.
• Globally installed skills will cascade and cannot be blocked at the project level, which will have to be clarified in the UI.
• A single list of skills is favored in the UI, with badges indicating whether each is installed globally or for the current project.
• Deleting a project will prompt a warning to the user and remove the associated symlinks.
• Dashboard views remain global.

The Interview Me process helps to clarify the product requirements and define a product spec. The presenter plans to feed the product spec back into Interview Me, focusing on the technical implementation (database schema, UI components, etc.). Finally, the entire output will be fed into "The Clanker" in plan mode, generating an execution plan to build the desired feature, which is meant to minimize wasted effort and keep the project on track. The speaker touts the ability to refine your thinking as the core value of the skill.