ATELIER — Design studio & adversarial UI review

Verify ATELIER's external tool dependencies are installed and reachable — a ✓/✗ table by tier (advisory; --strict to fail on a missing required tool).

Inspect the ATELIER plugin itself — audit skills, agents, and knowledge for drift, gaps, and duplication; produce a severity-ranked report.

Design a polished, design-reviewed UI mockup, wireframe, or user-flow — composed to the design canon and run through the convergent designer↔reviewer loop until it clears the fitness rubric. Produces commercial-grade material, not a first draft.

Adversarially review the design of any running SPA or screenshot — crawl the navigable routes (screenshot + accessibility snapshot per route), critique each against named design canon (Gestalt, UX laws, Nielsen's heuristics, WCAG 2.2), score it on the design-fitness rubric, and write a prioritised report. Asks which surfaces to review when ambiguous.

ATELIER INSPECTOR — on-demand agent that audits the ATELIER plugin it ships in (skills, agents, knowledge, commands under ${CLAUDE_PLUGIN_ROOT}). Triggered by user command only ("inspect ATELIER" / "/atelier:inspect"). Builds a fresh critical-analysis persona, reads every file, and produces ATELIER_INSPECTION_REPORT.md with severity-ranked findings (SUGGESTION / WARNING / CRITICAL).

ATELIER's heavyweight, SOTA-grounded adversarial design reviewer — the quality gate of the design studio. Spawned by the mockup skill to critique a rendered screen, a crawled SPA route, a screenshot, OR a generated/pictorial image (hero art, concept, illustration) against the named design canon (Gestalt, the UX laws, Nielsen's heuristics, Norman's emotional design, WCAG 2.2, and the art-direction canon — composition · light · colour · narrative · medium · the award bar), score it on the design-fitness rubric, and return prioritised findings that drive the convergent improvement loop. (Note: the ui-review skill performs inline critique for its own direct-review path; composed review by capability for other plugins.) Accepts an optional lens parameter to focus a pass: LAYOUT-REVIEWER (the legibility gate, runs first), HIERARCHY-REVIEWER, INTERACTION-REVIEWER, ACCESSIBILITY-REVIEWER, AESTHETICS-REVIEWER, CONSISTENCY-REVIEWER, or RICHNESS-MOTION-REVIEWER. Default is the full panel. Other plugins (e.g. PRESSROOM's image-aesthetic review) compose the AESTHETICS + RICHNESS-MOTION lenses by capability. Carries the KAIZEN self-improvement covenant.

Verify ATELIER's external tool dependencies are installed and reachable. Trigger with /atelier:check (or "check atelier prerequisites"). Runs a fast ✓/✗ probe grouped by tier; advisory by default, --strict to fail on a missing required tool. Reads the canonical manifest skills/check/requirements.tsv.

Design a polished UI mockup, wireframe, or user-flow — carefully composed and design-reviewed, not a first draft. Trigger with /mockup (or "mock up a screen for…", "design a wireframe", "draw the user flow", "sketch the onboarding screens"). Generates the artefact (renderable HTML/CSS screenshotted via the Playwright MCP, or an SVG wireframe; user-flows as Mermaid via pressroom by capability), then runs the convergent designer↔reviewer loop until it clears the design-fitness rubric. This is the capability IDEATOR calls so the user sees commercial-grade material, not whipped-up sketches.

Improve ATELIER itself by folding design feedback back into its canon and its design-fitness rubric — and by self-cleaving over-broad elements. Trigger with /atelier:self-improve (or "this design passed review but failed real users — fix the canon", "fold this design feedback in", "the reviewer keeps missing X"). Reflects one element against the covenant + pillars, applies the fix on a branch, runs the adversarial review (foundry's /foundry:pr-review if installed), and opens a PR so every future review and mockup, for all users, is sharper by default.

Adversarially review the design of any running app or screenshot — and write a scored, prioritised critique. Trigger with /ui-review (or "review the UI", "critique this design/screen", "design review", "what does the UI look like", "audit the accessibility of this page"). Crawls the navigable routes of an SPA via the Playwright MCP (screenshot + accessibility snapshot per route), or critiques a screenshot you paste. Grounds every finding in named design canon (Gestalt, the UX laws, Nielsen's heuristics, WCAG 2.2) and scores it on the design-fitness rubric. Use proactively whenever someone wants to know whether a UI is good — and why.

2 hooks across 1 event