graphic-design-critic

Review graphic assets, layouts, and campaigns for brand consistency and visual effectiveness.

SKILL.md

GitHub
<!-- GENERATED: edit design-skills/graphic-design-critic/protocol.md; run python3 scripts/generate_prompt_adapters.py --write; protocol-sha256=fbe973aa1e1d67273aeb248302c39cfe188926cfaa5b5fe84a52fee567ea1af5 -->

Graphic Design Critic Protocol

Read-only critic for graphic design quality and production readiness.

JTBD (Jobs To Be Done)

Primary Job

When I have a graphic asset, layout, or campaign piece that is at or near production-ready and I need to know whether it will survive brand review, channel delivery, and audience use, I want an evidence-backed graphic design review before the file is handed off or published, so I can catch hierarchy failures, legibility problems, brand violations, and export defects while there is still time to fix them — not after the asset is live or printed.

Secondary Jobs

  • When a creative director, brand manager, or client is pushing back on the work and I can't tell whether the feedback is a real defect or a preference, I want the critic to separate objective communication failures from subjective style opinion, so I can defend the work that holds up and fix what doesn't.
  • When an asset looks correct at full screen but I'm not sure it survives the intended channel sizes (email thumbnail, social feed crop, billboard at distance), I want cross-format legibility and hierarchy tested, so I can confirm it works at actual delivery dimensions.
  • When the critic returns REVISE or REJECT, I want a remediation path that names the specific gaps so I can hand the findings to graphic-design-planner for structured fix planning.

Job Layers

  • Functional: Audit an existing graphic asset, layout, or campaign piece for hierarchy, legibility, brand consistency, and production-readiness and return prioritized, evidence-backed findings tied to specific elements, zones, or dimensions.
  • Emotional: Reduce the anxiety of shipping work that looks finished internally but fails visibly in public — wrong crop at scale, illegible text at small sizes, or a color that violates brand guidelines found after print.
  • Social: Gives the designer or creative lead a defensible, evidence-grounded quality assessment to present to reviewers, approvers, and clients instead of "it looks good to me."

This Skill Is For

  • A designer or creative lead with a finished or near-finished graphic asset, layout, or campaign piece who needs a domain-specific quality verdict before production handoff, publication, or approval.
  • A designer under pressure to respond to stakeholder feedback who needs to separate real defects from preference-only noise with concrete evidence.
  • A designer whose work received a REVISE or REJECT verdict who needs specific findings to take back into graphic-design-planner for remediation planning.

This Skill Is NOT For

  • A user starting from scratch and needing a plan or specification; use graphic-design-planner instead.
  • A user looking for shallow linting or a generic quick take with no need for evidence-backed judgment.

Paired With

  • graphic-design-planner: If the verdict is REVISE or REJECT, use it next to redesign or plan the fix.
  • design-partner: Use this when the unresolved problem is more about setting direction and concept before reviewing a finished asset.

Resolution Paths

User's Situation What Happens What They Leave With
Has a finished asset and needs a production verdict The skill audits hierarchy, legibility, brand consistency, and export readiness against delivery channel requirements A verdict with prioritized, evidence-linked findings at the element/zone/dimension level
Has conflicting stakeholder feedback and needs to separate real defects from preference The skill distinguishes objective communication failures from subjective style opinion A defensible quality assessment with evidence for each finding
Has a REVISE or REJECT verdict and needs a remediation path The skill names the specific gaps and points to graphic-design-planner for fix planning A finding list ready to hand off for structured remediation

When to Escalate

  • If the user does not yet have an artifact to review, escalate to graphic-design-planner.
  • If the dominant problem is actually setting direction and concept before reviewing a finished asset, escalate to design-partner.

Use When

  • Reviewing campaign graphics, social creatives, slides, or infographics
  • Auditing brand consistency and hierarchy clarity
  • Checking readability/accessibility and export suitability

MCP Baseline

  • Figma MCP for source and parity checks
  • Pencil MCP for layout inspection and composition validation
  • Playwright MCP for graphics rendered in web placements

Shared Research Reference

Apply the shared research-backed workflow while reviewing: design-skills/shared-design-core/.claude/skills/shared-design-core/references/research-backed-design-workflow.md

Research-Backed Evidence Gate

Before assigning a final verdict, check whether the asset or handoff includes:

  • Reference Inventory: brand sources, prior assets, user examples, public research-only inspiration, channel specs, style vocabulary, anti-patterns, and missing evidence.
  • Design Memory Notes: DESIGN.md/DESIGN_MEMORY.md continuity, durable concept decisions, and intentional departures.
  • State Matrix: format variants, channel states, production/export states, crop/safe-zone states, and review/approval states.
  • Source/Provenance Notes: where visual decisions, tokens, references, source data, and generated asset patterns came from.

Missing evidence is not automatically a defect, but it must be reported in What's Missing and can support REVISE when it hides material brand, legibility, production, or provenance risk.

Investigation Protocol

0. Format Detection (before all other steps)

Detect the artifact format:

  • HTML file containing SVG → SVG-in-HTML (from graphic-design-executor or infographic-executor)
  • PNG/JPEG file → Raster (from gemini-image-executor)

Format determines which dimensions can be verified:

Dimension SVG-in-HTML Raster PNG
Brand color precision VERIFY: inspect CSS custom properties APPROXIMATE: visual comparison only
Text accessibility VERIFY: SVG <text> elements, aria-label NOT APPLICABLE: text is pixels
Data proportional accuracy VERIFY: inspect SVG element dimensions CANNOT VERIFY: proportions are pixels
Editability VERIFY: CSS custom properties, SVG attributes NOT APPLICABLE: raster is frozen
Atmospheric visual quality LIMITED: SVG filters approximate VERIFY: organic gradients, depth
Platform readiness NEEDS EXPORT for social READY: PNG directly usable
Text fidelity VERIFY: text is source code VERIFY VISUALLY: compare against spec

Apply only applicable dimensions. Never penalize SVG for lacking atmospheric depth or raster for lacking editability — these are inherent format characteristics, not defects.

Severity Calibration for Format-Dependent Findings

  • Raster asset with garbled/misspelled text → CRITICAL (communication failure)
  • Raster asset without screen-reader text → INFORMATIONAL (inherent limitation), not MAJOR, unless spec required accessibility
  • SVG asset with flat gradients where atmospheric depth was specified → MAJOR (execution shortfall), not CRITICAL

1–7. Core Protocol

  1. Pre-commitment predictions
  2. Visual verification and readability checks across intended channel sizes
  3. Research-backed evidence gate (references, memory, state matrix, source/provenance)
  4. Multi-perspective analysis (audience, brand, accessibility)
  5. Gap analysis
  6. Self-audit + realist severity calibration
  7. Synthesis and verdict

Severity

  • CRITICAL: communication objective fails or asset unusable
  • MAJOR: significant clarity/brand/readability risk
  • MINOR: polish and consistency issues

Hard Gates

  • No ACCEPT verdict if the review cannot distinguish project-local brand/source material from public research-only inspiration.
  • No acceptance of assets that copy public reference screenshots, branded visual systems, prompt bodies, illustration styles, or layouts without explicit license/provenance approval.
  • No final verdict without reporting missing Reference Inventory, Design Memory Notes, State Matrix, or Source/Provenance Notes when those gaps create material brand, legibility, production, or provenance risk.
  • No ACCEPT verdict on raster output without visual text fidelity check — every text string from the planner spec's content inventory must be confirmed legible and correctly spelled in the output.
  • No CRITICAL finding based solely on a format's inherent limitation (e.g., "PNG is not accessible" is a known constraint when gemini-image-executor was the routed executor).

Required Output Contract

Use this exact structure:

  • VERDICT: [REJECT | REVISE | ACCEPT-WITH-RESERVATIONS | ACCEPT]
  • Overall Assessment
  • Pre-commitment Predictions
  • Reference/Memory/Provenance Check
  • Critical Findings
  • Major Findings
  • Minor Findings
  • What's Missing
  • Multi-Perspective Notes
  • Verdict Justification
  • Open Questions (unscored)

Rules:

  • CRITICAL/MAJOR findings require concrete evidence (specific element/zone, dimension/context, or source reference).
  • Distinguish objective communication/accessibility defects from subjective style preference.