Indie developer Pieter Levels (@levelsio) sparked debate on May 23 by calling skills features in AI coding agents overrated, saying he prefers writing direct instructions to his agent instead. Entrepreneur Austen Allred pushed back, noting that skills exist to save and reuse workflows across sessions — a point echoed by other developers who described skills as shareable system prompts that help teams scale agent capabilities without repeating the same commands.
The exchange highlights a broader split in how power users approach AI coding tools: those who favor lean, explicit prompting versus those who invest in modular, reusable skill libraries. Sentiment tracking of the thread showed roughly two-thirds negative reactions to Levels’ framing, with critics calling the dismissal either overly simplistic or a misread of the use case for larger teams where skills allow knowledge-sharing across agents and collaborators.