agentleFS
PhlegonlabsPhlegonlabs/product-delivery-harness/skills/ui-design-builder/SKILL.md

ui-design-builder

Skill1 starsChanged 18 days agoMITView on GitHub

Use only when the user explicitly requests ui-design-builder or a requested skill workflow requires it. Turn an approved Product Definition into an owner-approved UI design package. Use after product-definition-builder for UI intake, direction studies, motion and media intent, frontend-design style integration, connected HiFi design-reference HTML, machine completeness review, Impeccable review, PRD-bound scoring, visual approval, and the Design System Need Gate. Do not use for product scope, backend architecture, technology-stack selection, or production implementation.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run only on the user's explicit request, continuation, or a required dependency of that requested skill workflow.

    line 8
  2. Distinctive:

    Read the owning section before its matching action.

    line 35
  3. Distinctive:

    Before UI research or authorship, read Required Skills And Inputs.

    line 37
  4. Distinctive:

    Before directions, HiFi or design repair, read the required inputs and Workflow.

    line 38
  5. Distinctive:

    Before handoff self-review, read Author Self-Review Before Handoff.

    line 39
  6. Distinctive:

    Before unresolved visual, motion, reference or platform decisions, read Reference Routing.

    line 40
  7. Distinctive:

    Before author dispatch, apply Frontend Visual Inputs.

    line 56
  8. Distinctive:

    Keep selected direction and approved HiFi identities traceable.

    line 59
  9. Distinctive:

    Keep copy source/completeness and PRD responsibilities valid.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

UI Design Builder

Run only on the user's explicit request, continuation, or a required dependency of that requested skill workflow. A matching task, installation, discussion, source edit or existing project artifact does not activate it. Required checks and dependencies stay within the requested outcome; a new top-level stage needs an explicit request.

Installed Commands

Resolve <ui-design-builder-skill-root> and logical script paths as Installed Commands states.

Purpose

At every invocation, run the entry checklist. A repair that restores the frozen design does not repeat initial design choices. Model the PRD's synthetic personas and loading, empty, permission, expired-session, failure, and recovery states. A mockup or HTML native projection is not real-login or native E2E evidence; that belongs to Harness.

Use this skill only after product-definition-builder has produced an owner-approved PRD.md, architecture.md, and stack-decisions.md, except for the owner-requested R0 exploration round below. Product Definition decides what the product does and which frontend, backend, data, auth, deployment, mobile, and commercial technologies it uses. This skill decides how the approved UI is structured and expressed.

For every UI-bearing product, this skill owns:

  • docs/design/ui-design.md;
  • direction studies under docs/design/directions/<round>/;
  • retained all-screens HiFi references under docs/design/ui-references/<run-id>/; and
  • the Style Intake, Motion and Media Intent, Direction Selection, HiFi Completeness Preflight, HiFi Review, Visual Approval, and Design System Need gates.

Existing wireframes and old design-system pairs remain inspection and validation evidence. They are not inputs to a new ui-design/3 package and never acquire new approval authority.

It does not change product scope, routes, actions, states, responsive targets, copy responsibilities, architecture, or stack by implication. Return those changes to product-definition-builder and resume only from a newly approved Product Definition revision.

Stage Routing

Read the owning section before its matching action. Domain and direction guidance stays conditional.

  • Before UI research or authorship, read Required Skills And Inputs.
  • Before directions, HiFi or design repair, read the required inputs and Workflow. Full initial authoring follows the complete required procedure in Workflow.
  • Before handoff self-review, read Author Self-Review Before Handoff.
  • Before unresolved visual, motion, reference or platform decisions, read Reference Routing.

Entry Boundary

  • A new round requires the approved Product Definition and Stack Decision Checkpoint, with the package's exact UI contract.
  • Agent configuration designates the UI author; this skill does not choose its agent type or model. The actual writer loads the complete pinned frontend-design skill; a parent read does not author design.
  • On the owner's request only, an optional R0 exploration round may run before Product Definition Approval: exploratory studies, no selection, no approval authority, no HiFi, no scope change. See Optional R0 Exploration Round.
  • Enhancements author only added or changed pages and necessary connecting flows. Routine maintenance does not rebuild historical HiFi or reopen frozen approvals.
  • Direction and HiFi authorship retain the required preflight, Impeccable critique/audit, browser evidence and H1-H9, then one human Visual Approval.

Design Lifecycle

Use references/review-workflow.md to classify this round. New initial design and explicit full redesign use ui-design/3; retained ui-design/2 packages keep their original sequence. Use the package selection rule in Required Skills And Inputs for the product preflight. Enhancements start from the current accepted product and author added/changed pages plus necessary connecting flows. Retain the accepted direction for unchanged enhancement scope; do not restart direction selection for unchanged decisions. Routine maintenance edits the actual product, effective requirements and existing change record directly; it does not require rebuilding HiFi or synchronizing historical tokens. An accepted product change is not blocked solely because historical HiFi differs. No historical decision is rewritten into a new approval. The PRD's ## At a Glance records the owner-selected Delivery profile: standard | mvp; Workflow states what mvp changes.

Review And Repair Boundaries

Before author dispatch, apply Frontend Visual Inputs. Direction exploration may keep selection pending; HiFi and its repairs carry the selected direction's captures, fonts, assets and concrete visual-feature sources.

  • Author directions and HiFi from the PRD contract, not from an implied page skeleton. The design author may choose hierarchy, grouping, section rhythm, typography, density and visual treatment; it may not add a page, route, control, state, responsive target, operation, copy obligation or integration.
  • Keep selected direction and approved HiFi identities traceable. A changed candidate needs applicable fresh evidence. The compiler writes a separate formal-pair preview; it never changes historical approvals.
  • frontend-design is the single design author for directions and HiFi. It owns visual-direction and frontend-authoring judgment, not contract compilation or implementation conformance. Impeccable critique/audit is a required review that gates Visual Approval; its authorization follows Impeccable Quality Review.
  • New machine observations use ui-evidence/3 without a human owner or attestation. ui-evidence/2 retains its historical human-attested semantics and is never fabricated or relabeled. Only the actual owner supplies direction and Visual Approval.
  • Keep copy source/completeness and PRD responsibilities valid. A changed product decision returns upstream; composition repairs inside accepted scope reuse authorization. Copy is reviewed with the full HiFi.
  • After the cheap completeness gate, each candidate gets one diagnostic wave of browser checks, Impeccable critique/audit and H1-H9. The rubric owns staffing, the first-wave axes, the thresholds and the repair budget. A candidate whose required checks stay unresolved stays blocked for Visual Approval, which remains the owner's call on the current candidate.
  • Numeric scores summarize quality; they never override a PRD contradiction, broken browser matrix, inaccessible required flow, dead control, or missing human approval.
  • A generated image or motion asset is optional unless the approved Motion and Media Intent record makes it required. Provider failure never authorizes a substitute treatment.

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.