matilha-design

name	matilha-design
description	Use when user is making UX, UI, or visual-design decisions in any phase — guides via core design heuristics and delegates to matilha-design-pack or matilha-ux-pack if installed.
category	matilha
version	1.0.0
optional_companions	["matilha-design-pack","matilha-ux-pack"]

When this fires

User asks "how should this look?", "is this UX good?", "what component to use?", or is about to commit design-level changes. Cross-phase; most useful in Phase 20/30 spec authoring and Phase 40 execution.

Preconditions

• None (works at any phase).

Execution Workflow

1. Pack detection: Inspect the ambient skill list for skills with plugin namespace matching matilha-*-pack (e.g., matilha-ux-pack:*, matilha-growth-pack:*, matilha-design-pack:* if installed). Group by plugin namespace. See skills/matilha-compose/SKILL.md for the canonical detection logic.

2. Intent classification: Does the user's design question touch cognitive load, error presentation, forms, accessibility, visual perception, growth-relevant moments (signup, paywall, activation), or broader design-system concerns? Use pack skill descriptions to inform the decision. Classify per pack: yes / no / partial.

3. Route based on intent:

   • Narrow intent (e.g., "which form component?", "what error format?"): invoke the single most relevant pack skill directly via the Skill tool. No preamble needed.
   • Exploratory intent (e.g., "how should this flow feel?", "is this UX good?"): if superpowers:brainstorming available AND ≥1 pack classified yes, build preamble per matilha-compose Step 4 template, emit it, then invoke brainstorming via Skill tool.
   • No packs installed OR packs irrelevant: apply core heuristics inline — simplicity (Krug Fato 1), recognition > recall (Weinschenk), progressive disclosure, consistent navigation, 5-rule error format.

4. Emit design guidance linked to the user's current task.

Rules: Do

• Prefer companion-pack skills when available (they carry neuroscience-backed depth).
• Cite sources in core heuristics (Weinschenk, Krug, wiki concept pages).
• Keep advice specific to the user's artifact.

Rules: Don't

• Make subjective "good/bad" claims without evidence.
• Override user's explicit aesthetic_direction in project-status.md.

Expected Behavior

Actionable design guidance. With companion pack installed, output quality is deeper; without companions, output is solid-core but less specialized.

Quality Gates

• Output references at least one heuristic (source cited).
• If companion pack invoked, mentions which one and why.

Companion Integration

Companion-pack awareness is handled via the pack detection + intent classification in steps 1–3 of the Execution Workflow. See skills/matilha-compose/SKILL.md for the canonical pack detection and preamble template.

• Any skill with plugin namespace matilha-*-pack (e.g., matilha-ux-pack:ux-*, matilha-ux-pack:cog-*, matilha-growth-pack:*, matilha-design-pack:* when installed) — detected at runtime via ambient skill list. Narrow-intent design questions invoke the most relevant pack skill directly; exploratory-intent questions emit a pack-aware preamble to superpowers:brainstorming.
• superpowers:brainstorming — invoked for exploratory-intent design questions (with pack-aware preamble emitted when ≥1 pack classifies yes).
• Otherwise: apply the core heuristics documented in Step 3 (simplicity, recognition > recall, progressive disclosure, consistency, 5-rule errors).

Output Artifacts

Usually chat output only; occasionally a commit to docs/matilha/design-decisions.md logging the decision with rationale.

Example Constraint Language

• Use "must" for: cite source for any design claim.
• Use "should" for: delegate to pack when available.
• Use "may" for: log decision in design-decisions.md if impactful.

Troubleshooting

• "no pack installed and user wants deep guidance": Recommend installing matilha-ux-pack or matilha-design-pack.
• "aesthetic_direction set but user wants to deviate": Log the deviation decision in recent_decisions via Edit.

CLI shortcut (optional)

No CLI equivalent (design guidance is inherently AI-conversational).

Sources

• [[concepts/leis-de-krug]]
• [[concepts/reservatorio-boa-vontade]]
• [[concepts/memoria-cognicao]]
• [[sources/100-things-every-designer]]