Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
[{"id":"calm_software_design_review.surface_audit_checklist","name":"Calm-Surface Audit Checklist","summary":"Checkable rules for screens and features — motion budget, surface count, attention cost, opinionated defaults, engagement-manufacturing red flags, notification posture, empty/loading/error states, hierarchy, onboarding, microcopy tone — plus the door test and the disappearance test.","when_to_load":["When reviewing any screen, feature, flow, notification channel, or copy surface — before producing surface findings.","When the user's concern is notification fatigue, streaks/gamification, motion overload, distraction, or anxious copy."],"path":"references/surface-audit-checklist.md"},{"id":"calm_software_design_review.simplicity_rubric","name":"Simplicity Rubric & Calm-vs-Delight Verdict","summary":"The judgment layer — foundational calm-school principles, Maeda's 10 Laws of Simplicity as an operational rubric (SHE, SLIP, Three Keys), the quality-without-a-name diagnostics, and the calm-school vs delight-school comparison table.","when_to_load":["When the surface checklist passes but the screen still feels off, or you need the subtract-the-obvious / sharpen-the-meaningful triage.","When rendering the calm-vs-delight verdict or deciding which school fits the product."],"path":"references/simplicity-rubric.md"},{"id":"calm_software_design_review.operations_and_roadmap","name":"Calm Operations & Roadmap Audit","summary":"Org-side checks — main-quest/side-quest discipline, default-no roadmaps, internal-MVP rule, 7-day zero-bug fix window, no per-feature engagement goals, written-async operations, cycles + cooldowns, org red flags.","when_to_load":["When reviewing a roadmap, feature list, team operating model, or per-feature metrics — before producing operations findings.","When triaging which features are main-quest, side-quest, or engagement manufacturing in disguise."],"path":"references/operations-and-roadmap.md"}]
The discipline this skill enforces is restraint instead of delight: calm software disappears during use, holds the user's attention only when it serves their work, and never manufactures engagement. Motion, surface count, attention cost, opinionated defaults, and notification posture are the load-bearing variables; confetti, streaks, badges, urgency timers, and celebration UX are failure modes. This is a procedure skill at domain altitude. The deep rules live in reference modules — the skill body holds the sequence, the output contract, and the escalation map. Load the references before producing findings.
Activation
Auditing a productivity tool, knowledge tool, dev tool, dashboard, agent workspace, or any product whose users arrive already cognitively loaded.
The user asks for restraint, calm, focus, fewer distractions, less notification fatigue, opinionated defaults, or quality in the Linear/Things/Basecamp sense.
A screen feels "off" but isn't broken — it checks every UX box and still feels like it's pulling at you.
Triaging a roadmap or feature list for main-quest vs side-quest vs engagement manufacturing in disguise.
Pre-launch review of a feature that may have drifted toward delight when calm was the brief.
Reviewing notification, email, or trial/billing messaging posture.
Judgment
Engagement-manufacturing findings outrank everything else — any single red flag is sufficient to flag the surface for rework.
Severity rubric:
high — engagement manufacturing (streaks, urgency timers, faux-FOMO, re-engagement nags, login rewards), push/email on by default, punishing absence, or prefers-reduced-motion ignored.
medium — motion or surface budget exceeded, configurability instead of opinion, demanding-not-inviting screens, anxious or mascot-voice copy, unrecoverable defaults.
low — tone polish, a single decorative flourish, minor density or whitespace issues.
Procedure
Ordered sequence and intent. Sibling-skill routing carries a → <id> marker resolved once in Routing; the three calm_software_design_review.* reference modules are this skill's own and are loaded inline.
Preflight. Confirm what is under review (screen, feature, flow, notifications, copy, roadmap, or operating model) and the user's arrival state. If the audience arrives under-stimulated — entertainment, social, consumer B2C — this is the wrong school: escalate → delightful_product_review and stop.
For any screen, feature, flow, notification, or copy review, load calm_software_design_review.surface_audit_checklist and walk categories 1–10 in order: motion budget, surface count, attention cost, defaults, engagement manufacturing, notification posture, states, hierarchy, onboarding, microcopy tone. (Precedence and severity are in Judgment.)
Close every surface with the checklist's door test and disappearance test. Subtraction moves come before additions in every fix.
If the checklist passes but the surface still feels off, or you need the subtract-vs-sharpen triage or the final verdict, load calm_software_design_review.simplicity_rubric and walk the Maeda Laws and the quality-without-a-name diagnostics.
For a roadmap, feature list, hiring plan, or team operating model, load calm_software_design_review.operations_and_roadmap and run the main-quest/side-quest question on every item plus the calm-operations checklist.
Tag out-of-scope findings for escalation instead of dropping them, routing each concern to its owner in Routing (→ ui_ux_quality_review, visual_craft_fundamentals, information_architecture_review, accessibility_inclusive_ui_review, delightful_product_review).
Assemble the report using the Contract below: findings ordered by severity then user impact, then the roll-up.
Routing
Out-of-scope concerns escalate to the owning sibling instead of being dropped (Procedure steps 1 and 6). One concern, one owner.
Out-of-scope concern
Escalate to
Audience arrives under-stimulated (entertainment, social, consumer B2C) — wrong school; or genuine delight opportunities for under-stimulated moments
delightful_product_review
hierarchy/spacing/type/color fundamentals (this skill assumes those already pass)
ui_ux_quality_review
works-but-lacks-polish craft execution
visual_craft_fundamentals
over-organization, navigation structure, or labeling confusion
information_architecture_review
semantics, focus, keyboard, or screen-reader behavior
accessibility_inclusive_ui_review
Contract
Every finding follows this canonical shape:
Category: <motion | surfaces | attention | defaults | engagement | notifications | states | hierarchy | onboarding | tone | operations>
Finding: <named rule violated, e.g. "Celebration animation on routine action">
Evidence: <specific screen, element, copy string, setting, email, or roadmap item>
Why it violates calm: <named principle or test — e.g. disappearance test, Maeda Law 7, main-quest discipline>
Severity: <high | medium | low>
Fix: <concrete change, subtraction moves first>
Delegated: <optional sibling skill id if the fix is out of this skill's scope>
Roll-up (after the findings list): top 5 highest-leverage fixes ordered by impact; engagement-manufacturing red flags found (state "none found" explicitly if zero); calm-operations risks if the org was in scope; a Law-10 audit (what is obvious → subtract, what is meaningful → sharpen); and a calm-vs-delight verdict — does this product fit the calm school, and where is it leaking toward delight?
Stop conditions before returning: every applicable category has at least one finding or an explicit "no issues"; every finding carries Evidence and Severity; the door test and disappearance test were run on each surface; out-of-scope concerns are tagged Delegated, not dropped.
Policy
A finding without evidence is not a finding — if it cannot be tied to a specific screen, element, copy string, setting, or roadmap item, do not include it.
Do not produce findings without loading the matching reference module first.
Do not recommend animations to "fix" calm. If the surface feels flat, the issue is likely emotion via voice, color, or detail (Maeda Law 7), not motion.
Do not add streaks, badges, or celebrations to "drive engagement" — manufactured engagement is leading-indicator churn in retention businesses.
Do not assume more configurability is better. Opinionated defaults are the move; configurability is what teams ship when they couldn't decide.
Do not push notifications by default, and do not manufacture urgency — no "limited time," no "only N left," no FOMO.
Do not confuse calm with sterile. Calm is restraint paired with strong opinion, not minimalism applied to soul — color, voice, texture, and signature interactions belong.
Do not strip features so aggressively the tool can't do real work. Calm ≠ feature-thin; if everything feels like a side quest, the discipline has lost its bias-to-ship.
Do not apply this skill to entertainment or under-stimulation apps — use delightful_product_review. Wrong school = wrong product.
Primary sources [PRIMARY]: Karri Saarinen (Linear — craft, door test, main quest), Werner Jainek (Cultured Code/Things — disappearance, do-not-over-organize), Jason Fried & DHH (37signals — calm-company doctrine), John Maeda (Laws of Simplicity), Steph Ango (Obsidian — constraints, file-over-app).
This skill is the on-brand counterweight to delightful_product_review; BuildOS surfaces default to the calm school. [internal-default]
Enrichment (2026-06-11): added four Saarinen/Linear org-side checks to operations-and-roadmap.md — quality-as-aspiration red flag (no owner/plan/rationale), principles-over-process / rigid-checkpoint-queue red flag, makers-in-direct-user-contact as the compensating control that makes no-A/B safe, and the organic-mentions qualitative KPI. Source [PRIMARY]: docs/research/youtube-library/analyses/2026-04-29_karri-saarinen-linear_craft-and-calm-software_analysis.md.
Maintainers: the canonical research draft with full lineage lives at docs/research/youtube-library/skill-drafts/calm-software-design-review/ (not available at runtime).
