Build a library of tested prompts that get consistent, system-aligned results from AI tools. Good prompts encode your system rules, patterns, and constraints so AI generates work that fits your system rather than fights it.
Prompt Libraries
How to
-
Identify common tasks
List where your team uses AI: generating component variations, writing documentation, creating specs, token naming, migration scripts.
-
Write task-specific prompts
Create prompts that include system context, constraints, and expected output format. Test and refine until they consistently produce good results.
- Example prompt: Generate a Button component spec following our system. Must use tokens from
color.action.*, include all states (default, hover, active, disabled), and meet WCAG AA contrast. Output as markdown with Props and Accessibility sections.
- Example prompt: Generate a Button component spec following our system. Must use tokens from
-
Include system rules
Embed key information in prompts: token naming patterns, accessibility requirements, brand constraints, and technical limitations.
-
Provide examples
Show AI good and bad examples of outputs. This improves consistency and helps AI understand nuanced requirements.
-
Organise for discovery
Make prompts easy to find: categorise by task, add search tags, and integrate into your workflow tools (Figma plugins, IDE snippets, Guidance setup).
-
Maintain and improve
Track which prompts work well and which don't. Update as your system evolves or as AI tools improve.