Lovable Launches Skills Feature to Turn Repeated Instructions Into Reusable Development Templates.

Lovable shipped Skills on May 18, a new feature that lets users save frequently used instructions as markdown files and reuse them across projects. The feature addresses a common friction point in AI-assisted development: having to re-explain the same patterns, preferences, and conventions every time you start a new build or chat session.

According to Tyler Bruno's explainer post, Skills work by storing instructions in markdown format within a project's /skills directory. When a user references a skill by name in chat, the AI reads the file and applies its instructions to the current task. The format is intentionally simple: plain English, written as if you're giving instructions to a junior developer.

Skills are markdown files saved in a project's /skills folder. Each file contains a set of instructions the AI should follow when building or modifying code. Users can create skills manually by writing a .md file, or by asking the AI to generate one based on a conversation.

Once saved, a skill can be invoked in any chat by referencing its filename. For example, a skill called error-handling.md might contain guidelines for how to structure try/catch blocks, log errors, and surface user-facing messages. When a user types "apply error-handling," the AI reads that file and follows its instructions.

The explainer emphasizes that good skills are specific, actionable, and focused on a single concern. Vague instructions like "make it modern" don't work well; detailed guidance on naming conventions, component structure, or API patterns do.

Bruno notes that the Skills format has spread rapidly across AI development tools. Cursor, Windsurf, Bolt, and other platforms now support similar markdown-based instruction files, often called "rules" or "prompts." The convergence suggests the format addresses a shared need: how to inject context and preferences into AI workflows without repeating yourself.

Lovable's implementation differs slightly in naming and storage location, but the underlying mechanic is the same. Users write instructions in natural language, save them as files, and reference them when needed. The format is portable; a skill written for Lovable could work in another tool with minimal modification.

The explainer includes guidelines for writing skills that actually improve output. Key recommendations:

• Be specific. Instead of "use good design," write "use 16px body text, 24px headings, and 8px grid spacing."
• Provide examples. Show the AI what you want, not just what you don't want.
• Keep it focused. One skill should cover one area: authentication, styling, error handling, API structure.
• Use imperative language. "Always validate user input before submitting" is clearer than "user input should be validated."

Skills can also reference external resources. A skill for an API integration might link to the API's documentation, or include example request/response payloads.

Skills address a problem that becomes more acute as AI coding tools mature. Early users often spend the first few messages of every session re-teaching the AI their preferences: how to name components, where to put files, what libraries to use, how to handle errors. Skills let you write those preferences once and apply them consistently.

The feature also supports team workflows. A shared skill file ensures that everyone working on a project follows the same patterns, even if they're prompting the AI in different ways. As we previously covered, Lovable has been moving toward more structured, repeatable workflows; Skills is another step in that direction.

Lovable's changelog also reveals two other updates shipped May 18: a new SEO and AI search tab with on-demand audits, Google Search Console setup, and Semrush-powered research, and design guidance improvements. The SEO tooling includes Lighthouse checks, sitemap validation, and structured data review, all surfaced in a dedicated Services tab.

Skills are available now to all Lovable users. The feature requires no additional setup; users can start creating .md files in the /skills directory immediately.