Bolt.new launches Skills: Make "configuration from zero every time" a thing of the past

Anyone who has used AI website building tools has had this experience: every time you create a new project, you have to say again "Whether you use React or Vue", "Which font to use" and "What is the code style?" The Skills feature launched by Bolt.new on July 22 is precisely aimed at this pain point - the official bluntly stated: "Every new project starts with the same configuration conversation, and Skills ends this repetition."

What are Skills?

Skills are a set of reusable "context, rules and workflow" bundles that cover your fonts, technology stack, code review standards, and more. Bolt.new automatically applies these rules at build time and can be added to a single project or shared across a workspace. For teams, this means that "team-level build specifications" can be defined once and take effect everywhere - when new members start the project, the AI ​​will already act according to team standards.

From "one-time generation" to "reusable specification"

Bolt.new is an AI full-stack application generation platform owned by StackBlitz. Users can generate runnable web applications using natural language. It made frequent moves in the first half of 2026: Bolt Slides (AI-generated presentations/interactive experiences) were released on July 16, and it was selected into Forrester's Agentic development platform landscape report on July 10. Skills allow Bolt to move from "one-time generation" to "reusable team-level build specifications" - this is essentially precipitating the team's engineering assets.

From an industry perspective, Skills/Rules/Workflows are becoming the standard capabilities of AI programming platforms: Windsurf's Skills, GitHub Copilot's AGENTS.md, and Claude Code's Skills are all doing the same thing - structuring the "team agreement" so that AI can act according to the rules from the beginning. Bolt.new provides an "ideal AI builder" for low-code/no-code users. The cleverness of its Skills design is that it does not require users to understand engineering. They only need to define "what style I want" and leave the rest to AI. For domestic AI website building/low-code tools, making "team norms" a first-class citizen is a key step to narrowing the experience gap.

Several directions worth tracking in the future:

1. Skills sharing mechanism: permissions and version management design for workspace sharing.
2. The emergence of skills market: Is there a third-party Skills template library for users to apply?
3. Echoes with Forrester’s selection: Whether Skills becomes a key piece of the puzzle for its “Agentic Platform” narrative.
4. Follow-up of domestic low-code platforms: Will AI website building tools launch similar "team specification" capabilities.