Browse the docs

Skills — Craft and build

/crisp-review

A grade in thirty seconds.

The core loopuser-invoked

A thirty-second design scan. /crisp-review checks a screen against all five CRISP dimensions, then returns a grade from A to F, the top three issues ranked by user impact, and a specific fix for each.

It is a diagnostic, not a report. It deliberately does not score each dimension individually — that is /crisp-audit's job. The output is built to be read in one glance and acted on in the next commit.

Reach for it when you need direction, not documentation:

  • Mid-iteration, between changes, when a full audit would break your flow
  • Before committing a screen — a fast pass to catch what you have stopped seeing
  • Grading agent-generated UI before you accept it
  • Deciding whether a surface is bad enough to deserve a full /crisp-audit
  • Grade A–F + one-line verdict
  • Top 3 issues ranked by impact
  • Specific fix per issue
  • Quick wins list
Claude Code — /crisp-review
B+
"Solid foundations. Three quick wins available before this ships."
P1Empty state needs a clear action CTA — users are stranded
P1Filter spinner blocks UI — switch to skeleton or optimistic update
P218 settings visible — only 4 are P0. Collapse with "Advanced settings"

Illustrative example output — not a real audit.

  • A — World-class. Ship it. Minor polish only.
  • B — Good. One or two fixable issues.
  • C — Functional but frustrating. Multiple P1s.
  • D — Users will struggle. Core experience broken.
  • F — Blocks users entirely. Do not ship.
What input does it need?

A screenshot, a component file, or a plain description of the screen. If .crisp.md exists it is loaded automatically, so the grade reflects your users and priorities rather than a generic standard.

Why only three issues?

Because ranked scarcity forces honesty about impact. A list of fifteen findings reads as noise mid-iteration; the three that most affect users, each with a specific fix, is what actually changes the next commit. Everything else lands under quick wins.

Will it soften the verdict?

No. The skill's tone rules are explicit: direct, specific, no softening. If a design fails, the review says it fails and exactly why — the goal is a better next iteration, not a comfortable one.

When should I upgrade to a full audit?

When the same surface keeps grading poorly, when you need severity ratings and a benchmark comparison to justify prioritisation, or when a milestone calls for the full scorecard. /crisp-review is the loop; /crisp-audit is the checkpoint.

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.