Design and review user flows, forms, dashboards, and onboarding for cognitive, sensory, and choice load — grounded in WCAG 2.2 cognitive-accessibility criteria and W3C COGA guidance, and applicable to every user, not only neurodivergent ones, because reduced cognitive load helps anyone under stress, time pressure, or divided attention. Use whenever designing or reviewing a flow with multiple steps, simultaneous choices, interruptions, or time pressure. Trigger on "too many options", "overwhelming", "cognitive load", "sensory overload", "ADHD/autism-friendly", "decision fatigue", "the user got lost", "too many steps", stacked modals, autoplay, or countdown timers. Not a visual-design skill (color/typography aesthetics) — it governs interaction patterns, information architecture, and pacing.
Accessibility framing treats cognitive load reduction as a feature for a
minority of users. The evidence doesn't support that framing: choice overload,
decision fatigue, and sensory overload degrade task completion for anyone
operating under stress, interruption, or divided attention — a neurotypical
user checking out on a phone while distracted is closer to a neurodivergent
user's baseline experience than either group's advocates usually credit.
Treat every rule below as a floor requirement for some users and a
quality improvement for everyone else, not as an optional add-on.
Composes with the library:
frontend-accessibility-by-construction — perceivability/operability
(WCAG Level A/AA structural requirements) is that skill's scope; this
skill covers pacing, load, and interruption, which WCAG 2.2's cognitive
criteria (COGA) also address but which are easy to satisfy technically
while still producing an overwhelming experience
resilient-ui-states — this skill's "never destroy user-entered work on
error" principle is the same rule that skill states from the resilience
angle; don't duplicate the implementation, cite it
MUST — no unrequested content change/interruption during input (autosubmit, auto-refresh)
WCAG 3.2.2 (A)
Forms, any input flow
User explicitly enabled the behavior
MUST — user can request more time or turn off a time limit, except real-time events
WCAG 2.2.1/2.2.6 (A/AAA)
Any session/task with a timer
The time limit is essential and real (e.g. a live auction) — must still warn and offer extension where technically possible
MUST — where a help mechanism (contact detail, self-help, chat, phone) appears on multiple pages of a set, it occurs in the same relative order/position on each, unless the user changed it
WCAG 3.2.6 Consistent Help (A) — verified against the exact SC text, 2026-09-22
Any multi-page or multi-step flow with a repeated help mechanism
N/A
SHOULD — plain, direct language: one point per sentence; a sentence using more than two clauses is a candidate to split
COGA, Making Content Usable, quoted directly: "Have one point per sentence" / "If a sentence can use more than two clauses, it typically can be broken up"
Legal/regulatory text with mandated wording — flag as an exception, don't silently rewrite
SHOULD — provide an undo/confirmation step before a destructive or hard-to-reverse action, instead of only a warning dialog
COGA / general usability heuristic
Delete, purchase, irreversible submit
The action is trivially reversible already (e.g. toggling a filter)
Provenance note, corrected 2026-09-22 (claim-provenance-discipline audit):
WCAG 3.2.6 Consistent Help does not specify a numeric interaction count to
reach help — an earlier draft of this skill stated "≤2 interactions" and
attributed it to WCAG/COGA; that number does not appear in the normative SC
text and has been removed. Reaching help quickly and without re-explaining
context is still good practice — it is now stated as a MAY-level heuristic in
the checklist below, with no numeric authority and no standards citation,
because none exists.
Empirical evidence: choice and cognitive load (not a W3C standard — behavioral research, and a section that needs its own provenance warning)
Provenance note, corrected 2026-09-22: an earlier draft of this section
claimed "capping simultaneous choices at roughly 5-9," implicitly Miller's
"7±2." That number is a documented UX myth in this exact form: Miller's Law
describes working-memory recall capacity (how many items a person can hold
in mind with nothing visible to check against), not how many visible,
scannable options an interface can show. A persistent on-screen menu relies
on recognition, not recall — NN/g and multiple UX sources explicitly warn
against capping nav items or visible choices at 7 for this reason. The
lawsofux.com source cited for the number does not, on inspection, state it.
The number is retracted; no numeric ceiling is asserted below.
Too many simultaneous options still measurably increase decision time,
anxiety, and abandonment (choice overload is real and well-documented) — but
the mechanism is decision load, not memory capacity, so the fix is
grouping/filtering/chunking a large option set, not hitting a specific
count. Apply this distinction:
Recognition context (a visible nav menu, a visible list of filters the
user scans rather than memorizes): there is no evidence-backed numeric cap.
Group by category when the list is long enough that scanning it becomes the
bottleneck — that is a UX judgment call per content, not a formula.
Recall/working-memory context (a set of options the user must hold in
mind to compare or decide between without re-reading — e.g. several
interdependent form fields that must be filled from memory of an earlier
step): this is where Miller's Law's actual claim (working-memory limits)
applies, and chunking into smaller groups is the correct mitigation.
Neurodivergent-specific guidance (secondary sources — practitioner literature, treat as heuristic, not standard)
Source: aggregated from ADHD/autism-focused UX design literature (practitioner
guides, not peer-reviewed standards — apply as heuristic, verify with real
users where the stakes are high).
Predictable layout: navigation and key controls stay in the same position
across the flow; nothing reorders based on state without the user
triggering it.
Optional low-stimulation mode: ability to disable non-essential animation,
autoplay media, and decorative motion without losing functionality.
Progressive disclosure: reveal detail only as needed, rather than
presenting the full depth of a flow/page at once — reduces the "freeze"
response to an unbounded amount of visible information.
Failure modes / anti-patterns
Fabricated urgency: countdown timers or "only 2 left" indicators that
don't correspond to a real, verifiable fact — this manufactures anxiety
without informing a decision, and is flatly incompatible with the "no time
pressure without user control" principle above. If the scarcity is real,
state it plainly; if it isn't, remove it.
Modal stacking: a confirmation dialog opening on top of another dialog
or a toast notification appearing while a modal is open — each interruption
compounds the working-memory cost of the one before it.
Silent context reset: a multi-step form or wizard that discards
progress on a back-navigation, browser refresh, or a minor validation
error — forces the user to reconstruct decisions they already made.
Ambiguous success/failure: an action that changes state with no visible
confirmation, leaving the user to guess whether it worked (did the "add to
cart" register? did the form actually submit?).
Choice-without-grouping: a filter panel or settings page listing 20+
ungrouped options flat, instead of chunked into named categories with
progressive disclosure.
Buried help: a support/contact path that requires navigating away from
the current task and re-explaining context already provided (e.g. re-typing
an order number already visible on screen).
Verifiable checks
No countdown timer, low-stock indicator, or "X people viewing this"
claim renders without being tied to a real, verifiable value.
At any point in a flow, at most one modal/dialog/toast requiring
attention is presented — no stacking.
A user can pause, leave, and resume a multi-step flow (refresh,
back-button, or explicit save) without losing entered data.
Any timed session either has no hard limit, or offers an extension/
warning before expiry (WCAG 2.2.1/2.2.6).
A visible/scannable list (nav, filters) has no numeric cap asserted
against it — check instead whether a long list (judgment call: long
enough that scanning it is itself the bottleneck) is grouped into
named categories rather than left flat. A recall-type set (options
the user must hold in mind without re-reading, not just scan) is
chunked into smaller groups — this is the case Miller's Law actually
covers; see the provenance note above before citing "7" for anything
else.
Every state-changing action (submit, add, delete, save) produces a
visible, unambiguous confirmation.
A help/contact path exists without requiring the user to leave the
task and re-supply context already given; if repeated across pages
within a set, it appears in the same relative position each time
(WCAG 3.2.6 — this is the actual normative requirement; no interaction
count is specified anywhere in the SC and none should be asserted as
if it were).
User-facing copy (errors, instructions, confirmations) is checked for
sentence length and subordinate-clause count — flag anything reading
as legal/technical register outside of actual legal text.
1---2name: non-overwhelming-ux3description: Design and review user flows, forms, dashboards, and onboarding for cognitive, sensory, and choice load — grounded in WCAG 2.2 cognitive-accessibility criteria and W3C COGA guidance, and applicable to every user, not only neurodivergent ones, because reduced cognitive load helps anyone under stress, time pressure, or divided attention. Use whenever designing or reviewing a flow with multiple steps, simultaneous choices, interruptions, or time pressure. Trigger on "too many options", "overwhelming", "cognitive load", "sensory overload", "ADHD/autism-friendly", "decision fatigue", "the user got lost", "too many steps", stacked modals, autoplay, or countdown timers. Not a visual-design skill (color/typography aesthetics) — it governs interaction patterns, information architecture, and pacing.4metadata:5 source: https://skillmd.com/skills/annatchijova/non-overwhelming-ux6---78# Non-Overwhelming UX910Accessibility framing treats cognitive load reduction as a feature for a11minority of users. The evidence doesn't support that framing: choice overload,12decision fatigue, and sensory overload degrade task completion for anyone13operating under stress, interruption, or divided attention — a neurotypical14user checking out on a phone while distracted is closer to a neurodivergent15user's baseline experience than either group's advocates usually credit.16Treat every rule below as a floor requirement for some users and a17quality improvement for everyone else, not as an optional add-on.1819Composes with the library:2021- **frontend-accessibility-by-construction** — perceivability/operability22 (WCAG Level A/AA structural requirements) is that skill's scope; this23 skill covers pacing, load, and interruption, which WCAG 2.2's cognitive24 criteria (COGA) also address but which are easy to satisfy technically25 while still producing an overwhelming experience26- **resilient-ui-states** — this skill's "never destroy user-entered work on27 error" principle is the same rule that skill states from the resilience28 angle; don't duplicate the implementation, cite it2930---3132## Standards (primary source: normative)3334Source: [WCAG 2.2](https://www.w3.org/TR/WCAG22/), [W3C COGA — Making35Content Usable](https://w3c.github.io/coga/extension/index.html), [W3C WAI36COGA task force](https://www.w3.org/WAI/about/groups/task-forces/coga/).3738| Rule | Level/source | Applies when | Does not apply when |39|---|---|---|---|40| **MUST** — no unrequested content change/interruption during input (autosubmit, auto-refresh) | WCAG 3.2.2 (A) | Forms, any input flow | User explicitly enabled the behavior |41| **MUST** — user can request more time or turn off a time limit, except real-time events | WCAG 2.2.1/2.2.6 (A/AAA) | Any session/task with a timer | The time limit is essential and real (e.g. a live auction) — must still warn and offer extension where technically possible |42| **MUST** — where a help mechanism (contact detail, self-help, chat, phone) appears on multiple pages of a set, it occurs in the same relative order/position on each, unless the user changed it | WCAG 3.2.6 Consistent Help (A) — verified against the exact SC text, 2026-09-22 | Any multi-page or multi-step flow with a repeated help mechanism | N/A |43| **SHOULD** — plain, direct language: one point per sentence; a sentence using more than two clauses is a candidate to split | COGA, *Making Content Usable*, quoted directly: "Have one point per sentence" / "If a sentence can use more than two clauses, it typically can be broken up" | User-facing copy: instructions, errors, confirmations | Legal/regulatory text with mandated wording — flag as an exception, don't silently rewrite |44| **SHOULD** — provide an undo/confirmation step before a destructive or hard-to-reverse action, instead of only a warning dialog | COGA / general usability heuristic | Delete, purchase, irreversible submit | The action is trivially reversible already (e.g. toggling a filter) |4546**Provenance note, corrected 2026-09-22 (claim-provenance-discipline audit):**47WCAG 3.2.6 Consistent Help does **not** specify a numeric interaction count to48reach help — an earlier draft of this skill stated "≤2 interactions" and49attributed it to WCAG/COGA; that number does not appear in the normative SC50text and has been removed. Reaching help quickly and without re-explaining51context is still good practice — it is now stated as a MAY-level heuristic in52the checklist below, with no numeric authority and no standards citation,53because none exists.5455## Empirical evidence: choice and cognitive load (not a W3C standard — behavioral research, and a section that needs its own provenance warning)5657Source: [Laws of UX — Choice Overload](https://lawsofux.com/choice-overload/)58(verified: does **not** state a numeric threshold), [NN/g and others on59Miller's Law misapplication](https://stephaniewalter.design/blog/your-menu-doesnt-need-millers-7-plus-minus-2-rule/),60[UX Myths — "Choices should always be limited to 7±2"](https://uxmyths.com/post/931925744/myth-23-choices-should-always-be-limited-to-seven).6162**Provenance note, corrected 2026-09-22:** an earlier draft of this section63claimed "capping simultaneous choices at roughly 5-9," implicitly Miller's64"7±2." That number is a documented UX myth in this exact form: Miller's Law65describes *working-memory recall capacity* (how many items a person can hold66in mind with nothing visible to check against), not how many *visible,67scannable* options an interface can show. A persistent on-screen menu relies68on recognition, not recall — NN/g and multiple UX sources explicitly warn69against capping nav items or visible choices at 7 for this reason. The70lawsofux.com source cited for the number does not, on inspection, state it.71The number is retracted; no numeric ceiling is asserted below.7273Too many simultaneous options still measurably increase decision time,74anxiety, and abandonment (choice overload is real and well-documented) — but75the mechanism is **decision load**, not **memory capacity**, so the fix is76**grouping/filtering/chunking a large option set**, not hitting a specific77count. Apply this distinction:7879- **Recognition context** (a visible nav menu, a visible list of filters the80 user scans rather than memorizes): there is no evidence-backed numeric cap.81 Group by category when the list is long enough that scanning it becomes the82 bottleneck — that is a UX judgment call per content, not a formula.83- **Recall/working-memory context** (a set of options the user must hold in84 mind to compare or decide between without re-reading — e.g. several85 interdependent form fields that must be filled from memory of an earlier86 step): this is where Miller's Law's actual claim (working-memory limits)87 applies, and chunking into smaller groups is the correct mitigation.8889## Neurodivergent-specific guidance (secondary sources — practitioner literature, treat as heuristic, not standard)9091Source: aggregated from ADHD/autism-focused UX design literature (practitioner92guides, not peer-reviewed standards — apply as heuristic, verify with real93users where the stakes are high).9495- Predictable layout: navigation and key controls stay in the same position96 across the flow; nothing reorders based on state without the user97 triggering it.98- Optional low-stimulation mode: ability to disable non-essential animation,99 autoplay media, and decorative motion without losing functionality.100- Progressive disclosure: reveal detail only as needed, rather than101 presenting the full depth of a flow/page at once — reduces the "freeze"102 response to an unbounded amount of visible information.103104## Failure modes / anti-patterns105106- **Fabricated urgency**: countdown timers or "only 2 left" indicators that107 don't correspond to a real, verifiable fact — this manufactures anxiety108 without informing a decision, and is flatly incompatible with the "no time109 pressure without user control" principle above. If the scarcity is real,110 state it plainly; if it isn't, remove it.111- **Modal stacking**: a confirmation dialog opening on top of another dialog112 or a toast notification appearing while a modal is open — each interruption113 compounds the working-memory cost of the one before it.114- **Silent context reset**: a multi-step form or wizard that discards115 progress on a back-navigation, browser refresh, or a minor validation116 error — forces the user to reconstruct decisions they already made.117- **Ambiguous success/failure**: an action that changes state with no visible118 confirmation, leaving the user to guess whether it worked (did the "add to119 cart" register? did the form actually submit?).120- **Choice-without-grouping**: a filter panel or settings page listing 20+121 ungrouped options flat, instead of chunked into named categories with122 progressive disclosure.123- **Buried help**: a support/contact path that requires navigating away from124 the current task and re-explaining context already provided (e.g. re-typing125 an order number already visible on screen).126127## Verifiable checks128129- [ ] No countdown timer, low-stock indicator, or "X people viewing this"130 claim renders without being tied to a real, verifiable value.131- [ ] At any point in a flow, at most one modal/dialog/toast requiring132 attention is presented — no stacking.133- [ ] A user can pause, leave, and resume a multi-step flow (refresh,134 back-button, or explicit save) without losing entered data.135- [ ] Any timed session either has no hard limit, or offers an extension/136 warning before expiry (WCAG 2.2.1/2.2.6).137- [ ] A visible/scannable list (nav, filters) has no numeric cap asserted138 against it — check instead whether a *long* list (judgment call: long139 enough that scanning it is itself the bottleneck) is grouped into140 named categories rather than left flat. A *recall*-type set (options141 the user must hold in mind without re-reading, not just scan) is142 chunked into smaller groups — this is the case Miller's Law actually143 covers; see the provenance note above before citing "7" for anything144 else.145- [ ] Every state-changing action (submit, add, delete, save) produces a146 visible, unambiguous confirmation.147- [ ] A help/contact path exists without requiring the user to leave the148 task and re-supply context already given; if repeated across pages149 within a set, it appears in the same relative position each time150 (WCAG 3.2.6 — this is the actual normative requirement; no interaction151 count is specified anywhere in the SC and none should be asserted as152 if it were).153- [ ] User-facing copy (errors, instructions, confirmations) is checked for154 sentence length and subordinate-clause count — flag anything reading155 as legal/technical register outside of actual legal text.
Run npx get-skill add annatchijova/non-overwhelming-ux in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Design and review user flows, forms, dashboards, and onboarding for cognitive, sensory, and choice load — grounded in WCAG 2.2 cognitive-accessibility criteria and W3C COGA guidance, and applicable to every user, not only neurodivergent ones, because reduced cognitive load helps anyone under stress, time pressure, or divided attention. Use whenever designing or reviewing a flow with multiple steps, simultaneous choices, interruptions, or time pressure. Trigger on "too many options", "overwhelming", "cognitive load", "sensory overload", "ADHD/autism-friendly", "decision fatigue", "the user got lost", "too many steps", stacked modals, autoplay, or countdown timers. Not a visual-design skill (color/typography aesthetics) — it governs interaction patterns, information architecture, and pacing. It is listed under Design & Media on SkillMD.
This skill has not completed SkillMD's automated safety review yet. Capability flags: makes network calls. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
annatchijova (@annatchijova) published this skill. Their other Agent Skills are listed on their SkillMD profile.