Browse the docs

Skills — Craft and build

/crisp-audit

The full evaluation.

The core loopuser-invoked

The full CRISP evaluation. /crisp-audit analyses a design with the critical eye of a senior product designer: a thirty-second first impression, then a scored pass through all five dimensions — Contextual, Responsive, Intelligent, Seamless, Powerful — with every violation named, rated P0–P3, and paired with a specific fix.

The result is a scorecard, a prioritised fix list, a benchmark comparison against Stripe, Linear, Notion, Asana, and Slack — or the benchmarks in your .crisp.md — and strategic recommendations for getting to world-class.

Reach for it at decision points, not during iteration:

  • Before a significant surface ships — the last structured look
  • Milestone reviews, where a grade and a scorecard need to justify themselves
  • Inheriting a UI you did not build and need to assess honestly
  • When /crisp-review keeps flagging the same surface and you need the full picture
  • Grade A–F with justification
  • 5-dimension CRISP scorecard
  • P0–P3 severity violations
  • Prioritised fix list
Claude Code — /crisp-audit
/crisp-audit settings.tsxGrade: B+
C
82Contextual
R
64Responsive
I
91Intelligent
S
71Seamless
P
88Powerful

Illustrative example output — not a real audit.

  • 30-second scan — first impressions, strengths, red flags, provisional grade
  • Dimension scoring — each of the five dimensions rated against explicit failure indicators
  • Severity rating — every violation classed P0 (blocks the user) to P3 (polish)
  • Benchmark comparison — where the design falls against world-class products
  • Action plan — fixes ordered by user impact, plus quick wins
How is this different from /crisp-review?

Depth and purpose. /crisp-review is a thirty-second diagnostic — one grade, top three issues. /crisp-audit scores every dimension individually, rates every violation for severity, benchmarks against named products, and ends with a plan. Review is for iteration; audit is for milestones.

What do the P0–P3 ratings mean?

P0 blocks the user entirely — an empty state with no recovery path. P1 is major friction the user can work around but should not have to. P2 is a noticeable degradation, and P3 is minor polish. The severity scale is what turns an audit from a list of opinions into an ordered plan.

What does it benchmark against?

Stripe, Linear, Notion, Asana, and Slack by default — each chosen for a specific strength, from Stripe's progressive disclosure to Linear's keyboard-native speed. If .crisp.md documents your own benchmark set, those take precedence.

Can it audit a screenshot?

Yes. A screenshot, a Figma link, or a written description all work. The richer the input, the more precise the violations — code or a live UI lets the audit point at exact elements rather than regions.

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.