| name | ux-remediation-and-redesign |
| description | Use when existing design-audit findings must become a prioritised remediation plan, redesign, re-test, and measured before/after result. Use design-audit or product-design-audit to diagnose problems and design-qa-and-pre-launch-review for the final release verdict. |
| metadata | {"portable":true,"category":"00-cross-cutting-ops-qa-a11y","compatible_with":["claude-code","codex"]} |
UX Remediation & Redesign
Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.
Use When
- You have a set of audit findings (severity-rated, ideally scored) for an existing,
already-built flow or product, and you now need to turn them into a sequenced, defensible fix
plan rather than a flat backlog.
- You need to prioritise competing fixes when you cannot do everything at once — to decide what
ships this sprint vs next quarter vs never, using a real method, not vibes.
- You are redesigning a broken flow and need the discipline that carries it from finding →
lo-fi → hi-fi → re-test → measured improvement, closing the loop back to the audit and forward
to the ship gate.
- You need to prove a fix actually worked — task-completion uplift, contrast now passing, a
cleared gate, a before/after KPI delta — not just assert "we fixed it."
Do Not Use When
- You don't yet have findings — there is nothing to remediate. Run
design-audit (a single
screen/artifact) or product-design-audit (a whole multi-platform product) first; this skill
consumes their output, it does not re-discover problems.
- You only need the diagnosis (the scored critique, the slop/WCAG/CWV gates) — that is
design-audit / product-design-audit. Do not re-run those gates here; cite their results.
- You need the final pass/fail go/no-go before launch — that is
design-qa-and-pre-launch-review. This skill hands a re-built, re-validated artifact to that
gate; it is not the gate itself.
- You are creating a new design from scratch (no existing flow to fix) — use the relevant
group 03–13 craft skills to build, then audit, then return here only if findings surface.
Required Inputs
| Input | Supplied by | Required? | Why |
|---|
| Severity-rated audit findings | Design or product audit | yes | Defines diagnosed problems |
| Baseline task and product metrics | Research/analytics owner | yes | Enables before-and-after proof |
| Constraints, red routes, and release window | Product and engineering | yes | Controls prioritisation |
- The audit findings — the severity-rated list from
design-audit / product-design-audit
(each finding: dimension, location, issue, impact, the standard it violates, a fix direction).
Without this, stop and run the audit first.
- The red routes — the product's critical paths (the few flows that, if broken, break the
business: sign-up, checkout, the core job-to-be-done). The triage cannot run without them; if
unstated, ask, then derive from the product's primary jobs-to-be-done.
- Effort signal — a rough design+dev cost per fix (a token swap is trivial; an IA
restructure is not). Get it from whoever will build it; estimate and label the estimate if not.
- The baseline — the pre-fix numbers you will measure against (task-completion %, time-on-task,
error rate, the audit score, the failing CWV/contrast values, funnel conversion). No baseline →
no provable improvement; capture it before touching anything.
- Constraints — release window, team capacity, any frozen surfaces, brand/legal limits.
Workflow
The remediation lifecycle is diagnose → triage → redesign → re-validate → measure, and it is a
loop, not a waterfall — each re-validation that still fails re-enters at triage. Full detail in
references/remediation-lifecycle.md; the prioritisation engine in references/triage-matrix.md.
-
Diagnose (intake the findings — do not re-audit). Ingest the audit's severity-rated
findings. Cluster them by root flow and by fix domain (hierarchy, conversion, a11y, IA,
content…) so related fixes ship together. Name each finding against the
doctrine/references/interaction-anti-patterns.md vocabulary (e.g. "Flat Hierarchy D1",
"Vague CTAs B3", "Silent System C2") — that gives every finding a known target right-pattern.
Carry forward which gate (slop / WCAG 2.2 AA / CWV) each finding feeds; gate-failing items
are pre-flagged because they cap the score until cleared (design-audit/references/audit-rubric.md).
-
Triage (sequence with the matrix — this is the core). Run every finding through the
three-lens stack in references/triage-matrix.md:
(a) Red-route severity (Travis): is it on a critical path? hard for users to overcome?
persistent/recurring? → Critical / Serious / Minor.
(b) Effort × Impact quadrant → fix the high-impact / low-effort quadrant first
("quick wins"), schedule high-impact/high-effort ("major projects"), defer the rest.
(c) MoSCoW → Must / Should / Could / Won't for this release.
Apply the two override rules: a gate-failing finding is always Must (a legal/charter
floor outranks the quadrant), and frequency-weighted issues (seen repeatedly in research,
not a one-off test artifact) outrank single sightings. Output a ranked fix queue.
-
Redesign (fix at rising fidelity, against doctrine). For each queued fix, design the target
state — conceptual → lo-fi → hi-fi — applying the right-pattern from
interaction-anti-patterns.md and the relevant craft skill (route via the table below). Every
redesign must satisfy the Anti-Slop Charter (doctrine/design-doctrine.md): state the
type/colour/layout choice, never reach for a banned default, prefer the authored option. A
fix that clears one finding but introduces a slop tell or a new gate failure is not a fix —
re-enter triage. Annotate each redesign with which finding(s) it closes.
-
Re-validate (prove it, don't assert it). Pick the validation method that matches the change
(references/remediation-lifecycle.md §Re-validate): 5-user usability test (task-completion %,
time-on-task, error rate — Nielsen's 5-user floor) for flow/hierarchy fixes; A/B test
(conversion / CTR uplift) for conversion fixes where traffic allows; tree test for IA
restructures; automated + keyboard + screen-reader re-check for a11y fixes (re-assert the
exact WCAG criterion now passes); field CWV re-measure for performance fixes. Success =
the task is completed without confusion and the failing standard now passes — not "users said
they liked it." A still-failing re-validation re-enters at triage, not redesign.
-
Measure (quantify the uplift, close the loop). Compare against the baseline from Required
Inputs: the audit-score delta (and any cap now lifted), task metrics (completion/time/
error before→after), product KPIs (funnel, bounce, session), and a summary instrument
(SUS / NPS / HEART task-success). Report the before→after as a table. Then close the loop:
hand the re-built, re-validated artifact to design-qa-and-pre-launch-review for the pass/fail
ship gate, and feed any newly surfaced findings back to design-audit.
Route each fix to its owning craft skill
| Finding domain (anti-pattern group) | Redesign with |
|---|
| Flat hierarchy, reading order (D1) | composition-and-visual-hierarchy |
| Typographic noise, banned fonts (D2) | font-selection-and-pairing, type-scale-and-spacing |
| Colour-only signalling, contrast (D3) | accessible-color-and-contrast |
| Icon soup (D4) | iconography-and-icon-systems |
| Form friction, forced sign-up, vague CTAs (B1–B3) | forms-and-input-design, ux-writing-and-microcopy |
| Idiot boxes, silent system, dead ends (C1–C3) | error-empty-and-system-messaging, micro-interactions-and-feedback |
| Deep nesting, mystery-meat nav (A2–A3) | information-architecture, navigation-design |
| Any a11y gate finding | accessibility-wcag-2-2-compliance |
| Any CWV / budget finding | performance-as-ux-and-core-web-vitals |
(Glob the engine for the exact skill name before citing; if none fits, name the doctrine
reference instead — never leave a fix un-routed.)
Anti-Patterns
- Re-auditing instead of remediating. This skill consumes findings; it does not re-run the
slop/WCAG/CWV gates. Re-deriving the diagnosis here duplicates
design-audit and wastes the
triage budget.
- Vibes-based prioritisation. "Let's fix the homepage first because it's visible" — without the
red-route × effort-impact × MoSCoW matrix the order is indefensible. Run the matrix.
- Skipping the gate override. Sequencing a contrast or banned-font fix as "Could" because it's
low-effort-low-impact-looking — a gate-failing finding is always Must; it caps the score
until cleared.
- Declaring victory without re-validation. "We redesigned it" is not "it works." No
task-completion re-test / cleared-criterion / KPI delta = not done.
- Fixing one finding and opening another. A redesign that clears the contrast finding but
introduces a slop gradient, or flattens IA but breaks a red route — re-enter triage, don't ship.
- No baseline. Measuring improvement you never benchmarked is unprovable. Capture the
before-numbers before the first edit.
- Treating the loop as a waterfall. Real remediation re-enters triage after each failed
re-validation; a one-pass plan that never re-tests is the book's exact anti-pattern.
Outputs
| Output | Consumer | Evidence / acceptance |
|---|
| Prioritised remediation queue | Product and delivery teams | Severity, effort, gate, rank, and owner |
| Annotated target-state design | Engineering and QA | Each change traces to a finding and craft rule |
| Re-validation and uplift report | Release owner | Before/after task, gate, and product evidence |
- A triaged fix queue — every finding ranked by the matrix (Critical/Serious/Minor ×
effort-impact quadrant × MoSCoW), gate-failing items forced to Must, with the sequence justified.
- A remediation plan — per fix: the finding(s) it closes, the target right-pattern, the owning
craft skill, the fidelity reached (lo-fi/hi-fi), and the chosen re-validation method.
- A re-validation record — the method run and its result (task-completion %, cleared WCAG
criterion, A/B uplift, re-measured CWV), with a pass/fail per fix.
- A before→after measurement table — audit-score delta (caps lifted), task metrics, product
KPI delta, and a summary instrument (SUS/NPS/HEART). This is the "fix worked" evidence.
- A handoff note to
design-qa-and-pre-launch-review (ship gate) plus any new findings routed
back to design-audit.
Examples
- Use the worked example in
examples/ for the audit-to-queue-to-revalidation lifecycle.
Quality Standards
- Preserve the original audit evidence and baseline before editing.
- Fix root causes along complete flows, then re-test the exact failed criterion or task.
- Block ship when a Must gate-failer remains unresolved or a redesign creates a new blocker.
Decision Rules
| Condition | Decision | Wrong-choice failure |
|---|
| Finding fails slop, WCAG, or performance gate | Make it a Must | Release blocker is deferred as polish |
| High-impact fix is low effort | Implement as a quick win | Best return waits behind major projects |
| Re-validation still fails | Return to triage and revise the approach | Failed fix receives cosmetic polish |
| Audit evidence or baseline is missing | Stop and obtain it | Improvement cannot be attributed or measured |
Capability Contract
Read and search are required; editing and execution require explicit remediation authority. Keep production mutation, research recruitment, and release decisions within their accountable owners. Preserve traceability from finding to fix.
Degraded Mode
If required evidence or tooling is unavailable, use the scoped fallback below and mark the result unverified.
Without editing, provide the ranked queue and annotated change specification. Without testing or baseline data, mark outcomes conditional and supply a re-validation plan; never claim uplift without evidence.
examples/checkout-remediation-worked.md — a full worked remediation of the Aurora Analytics
signup/landing flow whose flaws were diagnosed in design-audit/examples/design-audit-filled.md
(the 41/100, two-gates-failed audit). It takes those exact findings → triages them with the real
matrix (gate-failers forced to Must; one quick-win, one major project) → redesigns each against
the anti-pattern right-patterns and the Anti-Slop Charter → re-validates (5-user test + a11y
re-check + CWV re-measure) → measures the before→after (score 41 → 88, both caps lifted, task
completion 58% → 91%). Real numbers, no lorem, no banned fonts.
References
references/remediation-lifecycle.md — the five phases (diagnose → triage → redesign →
re-validate → measure) as a loop, with the per-phase entry/exit criteria and the re-validation
method-selection table.
references/triage-matrix.md — the concrete prioritisation engine: the Travis red-route
severity test, the effort × impact quadrant, MoSCoW, and the two override rules (gate-failers =
Must; frequency weighting), with a worked scoring table.
doctrine/references/interaction-anti-patterns.md — the diagnosis vocabulary and the target
right-pattern for each finding; every redesign aims at the "fix" column here.
doctrine/design-doctrine.md — the Mission and Anti-Slop Charter every redesign must satisfy; a
fix that introduces a slop tell is not a fix.
design-audit → references/audit-rubric.md — the upstream scored diagnosis this skill
consumes, and the gate/cap logic that makes gate-failing findings always-Must.
doctrine/references/wcag-2.2-criteria.md and doctrine/references/web-performance-budgets-2026.md
— the accessibility and performance floors a remediation must clear (and re-assert in re-validation).
Plugins (Load Alongside)
| Companion Skill | When to Load |
|---|
design-audit | Upstream — produces the findings this skill remediates (one artifact) |
product-design-audit | Upstream — produces product-wide, skill-routed findings to remediate |
design-qa-and-pre-launch-review | Downstream — the pass/fail ship gate the re-built artifact hands to |
composition-and-visual-hierarchy | When redesigning hierarchy / reading-order findings |
accessibility-wcag-2-2-compliance | When remediating any a11y gate finding (and re-validating it) |
performance-as-ux-and-core-web-vitals | When remediating any CWV / budget finding |
Sources: Lisandra Maioli, Fixing Bad UX Designs (Packt, 2018) — the diagnose→fix→measure
lifecycle, the Travis red-route severity test, the effort-impact matrix, and the re-validation +
measurement protocol — updated to 2026 and aligned to WCAG 2.2, Core Web Vitals, and the Chwezi
Anti-Slop Charter. Composes design-audit / product-design-audit (diagnose) upstream and
design-qa-and-pre-launch-review (ship gate) downstream. Right-patterns from
doctrine/references/interaction-anti-patterns.md.