Skill engineering and the case against one-shot AI design

Skill engineering is an emerging discipline focused on providing AI agents with the domain-specific vocabulary and context needed to perform complex tasks effectively. Paul Bakaus, creator of the open-source system Impeccable, argues that AI agents should not be used for one-shot, fully automated design. Instead, these tools should function as systems that allow humans to steer the creative process by defining precise operational meanings for subjective terms like bold, quiet, or dense.

Impeccable translates these design concepts into specific instructions regarding hierarchy, scale, and typography, ensuring that AI outputs align with existing design systems rather than producing generic results. By building these skills, Bakaus aims to bridge the gap between designers and engineers, allowing both groups to move up the stack and focus on high-level product decisions. This approach acknowledges that while AI can handle the initial 80 percent of a task, human judgment remains essential for the final 20 percent where taste and context are applied.

This philosophy rejects the trend toward total automation or fully autonomous agents. Bakaus maintains that there is no place for an auto-mode in his system because human involvement is necessary to provide purpose and maintain ownership over the final product. As design and engineering roles continue to converge, the focus shifts toward creating tools that empower users to articulate their intent clearly, ensuring that AI serves as a collaborative partner rather than a replacement for human creativity.