Browse the docs

Skills — Craft and build

/crisp-research

Start oriented.

Specialist extensionsuser-invoked

The research intelligence layer. /crisp-research takes a feature brief, classifies the feature type, and searches designated design sources — Mobbin, SaaSpo, Screenlane, Dribbble, and written case studies — for named patterns, mapped to the CRISP dimensions most at risk for that feature type.

The output is structured orientation, not inspiration: patterns worth studying, anti-patterns not to copy, benchmarks to beat, and the open questions the brief left unanswered. It feeds directly into /feature-design.

Reach for it at the start of feature work, before opening a single reference site yourself:

  • Starting a feature and about to lose an afternoon to Mobbin and Dribbble
  • A PM brief has landed and you suspect it has gaps — states, permissions, mobile
  • You want named, sourced patterns rather than a folder of screenshots
  • Feeding /feature-design with evidence instead of taste
  • Competitive pattern synthesis
  • CRISP dimension risk flags
  • Anti-patterns to avoid
  • Brief gaps surfaced before design
Claude Code — /crisp-research
topic: in-app approval flows
PATTERNLinear, Stripe both confirm inline — no redirect to a detail page
RISKModal-per-approval fails S — breaks flow on bulk actions
ANTIToast-only confirmation loses the audit trail users expect
→ feeds /feature-design

Illustrative example output — not a real audit.

  • No raw URLs or screenshot dumps — named patterns and named products only
  • No more than three patterns per dimension, ever
  • No padding when results are thin — it signals low confidence instead
  • No design recommendations — that is /feature-design's job
  • No designing into a monoculture — saturated aesthetic lanes are flagged so the design consciously departs from AI-generated defaults
What input does it need?

A feature name, a problem statement, or a pasted PM brief. If the brief is too vague to search on, it asks exactly one clarifying question and waits — it will not burn searches on a guess.

What is a saturated lane warning?

A flag that the brief points toward a visual monoculture — the editorial-typographic landing page, the generic SaaS cream dashboard, the glowing dark AI aesthetic. The warning names the lane and suggests departure directions so /feature-design makes a conscious choice.

How do the sources differ?

Mobbin for production UI patterns from shipped apps, SaaSpo for B2B and enterprise SaaS, Screenlane for micro-interactions, Dribbble for visual direction only — flagged as such — and web search for written case studies. The priority order changes by feature type.

What are the open questions in the output?

The specific gaps the brief left — states, permissions, mobile, triggers — surfaced as a maximum of three named questions for the PM. Every brief has gaps; the research names the ones that change the design.

zsh
$ npx skills add @laith-wallace/crisp

Installs all fourteen CRISP skills and auto-detects your AI harness. Run /crisp-teach once per project first.

The CRISP Letter

Design evaluation, in writing.

Occasional letters on making AI agents produce work worth shipping. New skills announced here first. No noise.