agentleFS
igovetigovet/cortex-orch-acp/skills/design-impeccable/SKILL.md

design-impeccable

Skill1 starsChanged 59 days agoMITView on GitHub

Critique, audit and refine an already-rendered product interface. Covers heuristic critique and technical audit (review, read-only) and targeted polish, bolder, quieter, distill, typeset, layout and clarify passes (refinement). Use when a visible interface exists and must be reviewed or improved within a named scope.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Use this skill for a surface that already renders.

    line 8
  2. Very specific:

    Read the packaged references through Cortex read_skill: design-critique for references/critique.md and design-refinement for …

    line 19
  3. Very specific:

    Follow design-critique for critique.

    line 27
  4. Distinctive:

    Keep factual copy: Ask before changing claims, numbers or product facts.

    line 60
  5. Very specific:

    Verify on the rendered result: Check desktop and mobile widths before and after the change, and open the captured images (see …

    line 64

≈ “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

Impeccable (Cortex adaptation)

Use this skill for a surface that already renders. It turns a visible interface into a sharper one through review and then targeted refinement. For a new visual direction, use design-frontend-design or design-ui-ux-pro-max.

Adapted for Cortex from the Apache-2.0 impeccable skill (see LICENSE, NOTICE.md and CORTEX-SOURCE.md). Its launcher, live browser mode, telemetry, persistence of critique snapshots and generated platform variants are not included.

Package references

Read the packaged references through Cortex read_skill: design-critique for references/critique.md and design-refinement for references/refinement.md. They are not files in the target project. Read only the relevant reference.

Role split

  • Review (read-only roles, such as ux_designer or accessibility_auditor): critique and audit. Produce findings and scores. Do not edit files. Follow design-critique for critique; audit reports findings by severity, with evidence, for the categories named in the table.
  • Refinement (write roles, such as frontend_dev): polish, bolder, quieter, distill, typeset, layout, clarify. Apply the smallest change inside the named target. Follow design-refinement.

A read-only role that finds a defect reports it with the evidence. A write role applies the fix and re-checks the same state.

Commands

CommandRoleUse whenOutput
critiquereviewYou need a design-director read: specificity, hierarchy, cognitive load, heuristics, personas.Findings with P0 to P3 severity and suggested command.
auditreviewYou need technical quality: accessibility, performance, theming, responsive behavior, implementation integrity.Findings by severity with the evidence for each.
polishrefinementA feature is functionally complete and needs a final consistency pass.Fixes for states, alignment, tokens, copy and cleanup.
bolderrefinementOne part of a surface reads flat next to its neighbors.One section raised to the expressive level the system already uses.
quieterrefinementA surface is too loud, saturated or busy.Restrained palette, weights, decoration and motion.
distillrefinementThe surface has more elements, copy or variants than its job needs.A simpler surface with the same functions.
typesetrefinementType hierarchy, scale, measure or pairing is weak.A type scale and roles applied consistently.
layoutrefinementSpacing rhythm, grid alignment or visual hierarchy is off.Spacing scale and grouping applied consistently.
clarifyrefinementLabels, errors, empty states or instructions are unclear.Plain-language copy that says what happened and what to do next.

Rules

  • Scope is sovereign. Touch only the named target. Do not restyle neighbors, add colors, fonts or primitives the surface does not already use, or migrate the page to a new idea.
  • The brief wins. A pinned aesthetic, era, palette or font is followed even when a general rule would prefer otherwise.
  • Refinement preserves; redesign replaces. Refinement keeps the current identity, behavior and copy. Replacing the visual world is a separate design-direction task.
  • Keep factual copy. Ask before changing claims, numbers or product facts.
  • Automated results are candidates. A detector or script finding is evidence of a possible defect, not proof of quality, and a clean scan does not replace visual judgment.
  • Verify on the rendered result. Check desktop and mobile widths before and after the change, and open the captured images (see design-visual-qa).
  • No self-scoring as evidence. Scores in a critique are judgments that must cite the observed element. They do not certify quality by themselves.

Refinement procedure

  1. Establish the system: project design documentation, tokens, shared components. Classify each drift as a missing token, a one-off implementation, a conceptual mismatch or a local defect. Fix it at the narrowest correct level.
  2. Gather evidence: render the target at desktop and mobile widths and exercise its loading, empty, error, success and disabled states.
  3. Triage: broken or blocked tasks and misleading state first, then missing states, then flow, hierarchy and responsive drift, then visual and motion inconsistencies, then cleanup.
  4. Apply the change, remove dead code and temporary artifacts, and keep the diff limited to the target.
  5. Re-check the same states and viewports, open the screenshots, and report what changed and what was not verified.

Output

Name the command, the target, the evidence (paths of opened screenshots and the states exercised), the findings or changes with their severity, the files changed, and the checks not run with their reasons.

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.