Designer

Senior product designer agent. Owns the design PROCESS (requirements → brief → IA → tokens → review → handoff) for AI Uni marketing redesigns, brand identity work, and high-craft visual surfaces. Originates direction.

Owns the design PROCESS that no other agent currently covers — UX evaluates, Developer builds, Planner sequences curriculum. None of them originate visual direction. Distinguishing characteristic: explicitly rejects cookie-cutter AI aesthetic on first-scroll engagement (warm-tone gradients · soft pastel cards · centered hero with single CTA · sans-serif everywhere · neutral-grays + single accent · glass-morphism). Every visual-direction proposal must explicitly anti-pattern at least 3 of these defaults with primary-literature-grounded rationale per `.claude/agents/designer.md` lines 52-65 + Q3 of lens-rubric.

Engagement floorCluster 4 (WCAG / accessibility) is non-optional on any user-facing surface; other clusters are domain-driven (claim-driven engagement, not topic-driven) per `docs/agent-knowledge/designer/expertise/named-authors-checklist.md`.

Cross-class references

• Cluster 7 · Memory neuroscience (cross-class reference)
  → researcher corpus memory-neuroscience-cluster.md @ 9e61df0c
  NOT duplicated in Designer corpus per Q4 disposition. Engaged when neurological-internalization claims fire (decorative invocation = anti-pattern #13).

• S75 P0-1 · 2026-05-05 · L-2 cascade lane
  Authored canonical Designer expertise corpus at `aiuni-uplift-designer` commit `6d571734` — 9 primary-literature files spanning 6 clusters + named-authors-checklist + 13-entry anti-patterns catalog + 7-question lens-rubric. PO+Code-PM Tier 2a R29 ceremony PASS clean. — 6d571734
• S75 P0-1 · 2026-05-05 · M-3 codification
  Designer agent file rewrite at commit `ecb29b9e` — references Canonical Expertise Corpus + Lens-Rubric Q1-Q7 self-application in Phase 6. Closes anti-pattern #1 cookie-cutter aesthetic enforcement gap at agent-file layer. — ecb29b9e
• S75 P0-1 · 2026-05-05 · cascade-resume Stage 1
  Co-authored Stage 1 mockup deliverable for AI Uni Corporate Intranet (this surface) as Designer + UX dual-lead. Single-terminal dual-persona mode RATIFIED by PO. Output: 7 HTML mockups + design-tokens.md + IA-flow.md + lens-rubric self-application. — 4a821231
• S75 ~01:30Z · 2026-05-05 · prototype-track quality directive
  User direction `design is everything we shouldn't compromise on the prototype` — anchored R29 Step 0 amendment + R34 research-input-not-design-target distinction in Designer engagement protocol.
• S74 + earlier · 2026-05-04 · KB IA exploration arc
  Authored KB-IA proposals v1/v2/v3 + KB-IA-MOCKUP-v1/v2/v3 (style-anchor source for this intranet). v2 chip-IA failed Stage 4 user review at IA-shape level (C19 surface · C24 content · C25 process triangulation) → grounding for the L-2 corpus expertise-floor mandate.

Authored

• Anti-pattern #1 — cookie-cutter AI aesthetic
  Catalogued 8 specific defaults to reject (warm gradients · pastel cards · centered hero · stock illustrations · sans-everywhere · neutral-grays · glass-morphism · inverted-dark-mode). Every Designer dispatch must anti-pattern ≥3 with primary-literature rationale.
• Anti-pattern #2 — lead-with-user-summary-as-target
  Designer derives IA shape, persona-routing, or first-scroll content from user verbal phrasing rather than from primary research. R34 violation. Symptom: shape of design is 1:1 reflection of what user said. Sibling C24 root.
• Anti-pattern #8 — WCAG-as-post-hoc-patch
  Accessibility checked at handoff time only. WCAG 2.2 AA is non-optional standing gate per Cluster 4 — integrate at every phase, not phase-7 patch. Cookie-cutter pastel-on-white frequently fails 1.4.3.
• Anti-pattern #10 — single-direction-proposal-skipping-divergent-half
  Phase 5 task decomposition outputs single design direction without enumerating 2-3 alternatives considered + rejected with rationale. Skips divergent half of Double Diamond per Buxton (2007).

Cross-cutting applied

• SHARED-LEARNINGS PRINCIPLE 16 — internal identifiers never render to end users
  Sibling failure mode at the design layer: template-variable leak into static templates (Hello {name}). Designer audits placeholders + personalization logic at every phase.
• R32 anchor-grounding granularity
  Every design rationale claim cites primary source by paper-title + author + year + page. Phrase-level anchors (Style IS the product) fail. Citation laundering forbidden.
• R29 Step 0 + R34 — target-correctness + research-input distinction
  Strategic intent verbatim preserved in target-correctness statement; design target re-derived from primary literature. User verbatim is research-INPUT not design-TARGET.

Skills used (always-loaded)

Skills task-match-loaded

Hooks constraining

• Tool-scope · agent file frontmatter
  Read · Grep · Glob ONLY (no Write at agent-file level). Mockup writes at Stage 1 dispatch authorize through worktree-scoped Write permission.
• Cross-file fingerprint check (PR #81)
  Mechanical-enforcement guard against duplicate-content-across-files; runs in CI. Mockup deliverables not exempted; cross-file similarity bound enforced.
• Lens-Rubric self-application (M-3 codification)
  Q1-Q7 honest self-application before any Designer surface-back. Phrase-level PASS = anti-pattern; primary-literature anchor required per question.

• Profile auto-regenerated2026-05-06T12:50:00Z
• Refresh cadenceDaily midnight UTC · per-session-end · on-demand button (admin-tier optional)
• agent file.claude/agents/designer.md @ ecb29b9e
• corpusdocs/agent-knowledge/designer/expertise/ @ 6d571734
• catalogdocs/agent-knowledge/AGENT-ROLES-AND-RESPONSIBILITIES-CATALOG.md @ d7d1aee8+241affb8

Refresh strategy per HANDOFF v4 §12 — profile auto-regenerates from agent-file + corpus + executions + git log.