All products

Document design projects with AI

Keep project rules, implemented design values and current progress in a small set of documents that coding agents can read and maintain.

Document design projects with AI through records that describe the code as it exists. The working-docs skill separates operating rules in AGENTS.md, implemented visual values in DESIGN.md and progress, decisions and next steps in CONTEXT.md.

Unknown design values stay unknown instead of becoming invented tokens. Unconfirmed content and unperformed checks remain visible. The documents help carry project decisions between sessions, but reading them or writing a checklist does not replace running the actual checks and inspecting the result.

What it does

  • Record project rules and approval boundaries
  • Describe implemented tokens and components
  • Maintain dated progress and next steps
  • Prepare context for the next coding session
  • Preserve decision reasons and recurring traps
  • Split long records into useful working documents

Install and use

  1. Load the instruction and licence notice in a file-capable coding agent. Provide the repository, existing rules, token source and current design decisions; verification dependencies come from the actual project.
  2. Ask: “Create working documents for our booking site. Preserve existing colours, mark undecided design values and require my approval for deployment.”
  3. Review the rules, design and context drafts. Check command names, file paths and the reasons behind restrictions. Planned design must not be recorded as implemented.
  4. Update context after each change and revise the design record when tokens or motion change. State which checks ran and what remains unfinished; commits and publication follow the project’s own rules.

FAQ

Source

Made by UV. Original skill: working-docs, published on GitHub under the MIT licence.

View on GitHub