agentleFS
ArcanEdge-AIArcanEdge-AI/coding-agent-playbook-codex/skills/taste-new-website/SKILL.md

taste-new-website

Skill3 starsChanged 3 months agoMITView on GitHub

Use when the user wants Taste's staged workflow for a new website, including brief questions or project autofill, design approval, a Next.js page, and written design audits.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the invocation selects auto, fill the brief from the current project.

    line 10
  2. Distinctive:

    Use any additional invocation text or answers already given in this chat.

    line 15
  3. Distinctive:

    Avoid: specific patterns or visual choices to avoid, or none?

    line 24
  4. Distinctive:

    Stop and request the user's approval of this direction.

    line 34

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Run the new-website workflow from the Taste Skill prompt guide. Read the sibling design-taste-frontend skill for the installed Taste release's design rules and commands. Resolve <skill> in those commands to that installed skill's folder, and run them from the project root. If the dependency is missing, report it before starting. Follow governing instructions and the user's project constraints.

Collect the brief

If the invocation selects auto, fill the brief from the current project. If it selects questions, ask the brief questions. Otherwise, first offer these two choices and wait for the user's selection:

  • Fill the brief from this project.
  • Answer the brief questions myself.

Use any additional invocation text or answers already given in this chat. Do not repeat questions that are already answered. Use the available question UI when supported; otherwise ask in chat and wait.

Collect these six fields:

  1. Page kind: landing page, portfolio, or marketing site?
  2. Product: its name and a one-sentence description?
  3. Audience: who will read or use it?
  4. Vibe: two to four concrete adjectives, such as minimalist, editorial, or restrained?
  5. References: actual URLs or product names that anchor the aesthetic, or none?
  6. Avoid: specific patterns or visual choices to avoid, or none?

For auto, inspect relevant read-only project evidence: PRODUCT.md, DESIGN.md, README, package and framework files, existing routes, visible copy, components, and public brand assets. Populate all six fields where the evidence supports them. Distinguish documented facts from inferred suggestions and unknowns. Ask only about material gaps or conflicting evidence, together in one concise follow-up. Do not invent product claims, customers, references, or brand facts. An empty project still needs a brief from the user.

Once intake is complete, show the filled brief with any assumptions, then enter Step 1. Filling the brief does not approve implementation.

Step 1: Design read

Use Taste's “Read the brief and the project,” “Write the argument,” and “Draw the direction” steps. Deal options with the sibling skill's draw command and --options 3, using the brief's sector and section roles. Present the proposed option ID, a one-sentence design direction, and why it fits the audience and content.

Stop and request the user's approval of this direction. Do not create or edit the website before that approval. If the user changes the brief or direction, incorporate the change before proceeding.

Step 2: Build after approval

Commit the approved option with the same draw arguments and --direction <chosen id> in place of --options. Build one Next.js page with at least eight sections that fit the product and at least four layout families. Follow the resulting direction sheet and Taste's “Build it, still” and “Move” steps. Use project imagery or authorized, credited assets according to Taste's imagery guidance. Keep one consistent theme across the page.

Step 3: Verify

Report each audit in writing:

  • Dash audit: zero U+2014 em dashes or U+2013 en dashes in the produced content.
  • Follow Taste's “Gate” and “Look” steps. Lint the source and inspect the production build at the default 1440, 768, and 390 pixel widths. Run the scene check when the page uses scroll scenes. Open the screenshots, write the critique using Taste's rubric, and record inspector failures or warnings and their disposition. Repeat affected checks after fixes and after adding motion.
  • Section layout repetition: list the layout family of every section.
  • Hero discipline: check headline line count, subtext word count, and whether the primary CTA is visible without scrolling.

Any Fail prevents completion. Correct in-scope failures and rerun the affected checks, or report the specific blocker. Do not mark an unverified item Pass.

Source: Taste Skill prompt guide, captured 2026-10-06. This skill adds interactive intake and optional project autofill to the guide's new-website workflow.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.