# Experience Remediation

> Execute evidence-backed UI and UX remediation using project layout contracts, shared components and complete OJO/Taste/Impeccable methods, then supply per-rule evidence for independent acceptance.

- Skill: `shuaimxu/experience-remediation` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx get-skill add shuaimxu/experience-remediation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/shuaimxu/experience-remediation/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: ShuaiMXu (https://skillmd.com/u/shuaimxu)
- Updated: 2026-11-16
- Page: https://skillmd.com/skills/shuaimxu/experience-remediation

---


# Experience remediation

Use for a confirmed experience repair, including AI slop and inconsistent shared components. The task, adopted project rules and approved Foundation take precedence over generic examples. This Skill never authorizes unrelated redesign, publication or a human acceptance decision.

## Diagnose

Read the complete assigned method documents. Start with the current page and its actual working viewport. First review the recurring AI slop and UI details: dropdown arrows/checkmarks, selected/focus states, gaps, boundaries, radii, redundant navigation and unnecessary copy. Then inspect shared components, task flow and current-page hierarchy/grid. Phone and other viewport checks come last as regression coverage; they never replace current-page remediation. For every contract rule and supplied state, return an explicit result with original screenshot references. Distinguish a visible violation from a preference, missing evidence and a genuine non-applicable rule. Never label an unchecked state not-applicable. Follow the existing finding-admission policy when proposing a repair.

## Mandatory order

Current-page AI slop and detail review → shared-component correction and current-page interactions → task/hierarchy/layout validation → cross-device regression. Inspect historical project issues before introducing new aesthetic preferences. Group findings and verification in this order, even if captures were collected together. Do not declare success because a mobile overflow was fixed while current-page control defects remain.

## Decide and implement

1. Reproduce the issue and identify the user task it interrupts. State what to preserve.
2. Inspect the actual design sources, shared component and all known consumers. A package dependency does not establish actual use.
3. Record a concrete decision: content order, grouping, primary action, copy retained/removed, widths, spacing, responsive reflow and state behavior. Use existing tokens and primitives. Resolve a shared cause before patching instances.
4. For grids, inspect the adopted container width, columns, gutter, outer margins, alignment and breakpoints. Record actual source references and the chosen values/reflow in the result. Preserve intended reading order and nested alignment; never impose OJO's 12-column/8pt examples on a different approved grid. Missing layout evidence must be stated, not invented.
5. Use OJO component-recipe for applicable control states and layout patterns; visual-tokens for spacing and alignment; icon-guidelines for arrows and indicators; motion-system for feedback and reduced motion. Read relevant complete references. Follow the selected Impeccable specialist in the bound implementation readFirst route, then Polish (or Polish alone), with web-taste's applicable composition guidance, constrained by its Harness adapter and the actual surface. Do not load unrelated specialists by default. Keep all original packages intact.
6. Check default, hover, pressed, selected, open, focus, disabled and relevant loading/error/success states. Do not invent asynchronous states for static content. For dropdowns check long labels, chevrons, checkmarks, item gaps, menu boundaries, scrolling, keyboard selection, Escape and focus return.
7. Write the structured execution result required by the Harness. Every rule and delegated issue must have one result; fixed items name actual changed files. Preserved and false-positive decisions require evidence and a reason. Blocked items stay visible. Reading a method is not completion.

## Page structure and original-image review

`UX-page-structure` requires explicit grid, spacing, alignment, component-usage and responsive-state review for each assigned original. Mark supported defects on the original capture with normalized bounds linked to the finding ID. Keep annotations as a derived overlay; never alter or replace original evidence. Use actual target-product layout sources, not mandatory universal 8px spacing or arbitrary column counts. Missing/unobserved dimensions remain needs-evidence and block pass. Follow the structured `structureChecks`/`annotations` protocol in [page structure review](../../docs/page-structure-review.md).

## Product icon hard gate

`UX-product-icons` is required for every newly staged strict project/HappyHands remediation contract. “Codex default icon” means an unexamined generator fallback, not a maintained blacklist of icon libraries. First decide whether an icon is necessary. Omit or remove unnecessary decoration; text-only/no-icon output is valid, and no icon may be added just to satisfy this gate. For necessary icons, preserve already suitable existing choices with evidence rather than forcing redesign. Never reuse default glyphs just because they compile. Establish the target product's icon family and source, map icons to its actions and vocabulary, and inspect stroke/fill, optical size, spacing, states and accessibility. Existing assets or a standard library are acceptable only when deliberately fitted to this product; custom drawing is not required for familiar functional symbols. Do not copy the HappyHands shell's icon language into third-party products.

Execution must record concrete source/adoption evidence in `observed`, the product-specific adaptation in `decision`, affected source files and actual checks in `verification`. Independent review must cite each assigned state's original and explain the observed semantic and visual fit in its `reason`, not repeat the executor's claim. Missing evidence stays blocked/needs-evidence. A no-icon scope, including deliberate removal of unnecessary icons, may use not-applicable with a concrete necessity explanation. No automatic claim of originality, correct semantics or human approval is made by this protocol. Old run snapshots remain unchanged; new runs without strict remediation are not covered by this gate.

## HappyHands homepage Hero

When the bound profile includes `surface: homepage-hero`, apply the two Hero rules to the actual HappyHands portal homepage states. Treat the existing cover spiral as the approved motion exception. Keep the archive covers, orbit, center clearance, and existing content and actions. Check perspective, opening tilt, pointer response and scroll exit using timed frames and actual browser evidence. Examine the glass header while covers pass behind it. The central glass mark has one silhouette with the highlight clipped inside it. Check both themes and desktop and phone. Reduced motion freezes the covers, WebGL failure uses the cover fallback, and the offscreen animation stops. Record unobserved motion as needs-evidence; the bound `hero-motion` specialist check is required for a passing review.

## Independent acceptance

The reviewer does not edit. Inspect originals and supplied verification records, not only the executor's account. Check every rule on every assigned state, including layout/grid relationships and shared consumers. A pass requires complete coverage with no unresolved failures or missing evidence. Record limitations and exceptions; never infer unseen interaction behavior from screenshots. Preserve existing source/hash, visual QA and human-judgment gates.

The optional `harness compare` export presents eligible viewport originals at 1:1 with provenance. Reviewers must still inspect the original images, task and actual interaction evidence. An exported image cannot turn a failed or missing review into a pass. Missing capture-time binding requires fresh capture, not retrofitted hashes or edited records; a DSH final event is likewise not independent acceptance.

## Evolving these methods

Keep project preferences scoped to that project. New general rules require an accepted real case, applicability and exceptions, a regression example and an independently reviewed version. Never rewrite historical contracts when updating this Skill.


## Required website and landing-page method

For official/company websites and landing pages, web-taste is the required primary design and implementation method. Classify these tasks as public-page before formal diagnosis; read its complete copy, voice, system, layout and section references and applicable asset/motion references. OJO and Impeccable supplement this execution; they do not replace web-taste. Preserve user-approved project rules and evidence boundaries. Do not silently route website/landing-page work through a legacy Taste profile.


## Required task and design handoff

For new lightweight reviews, follow the complete [experience orchestration](../../docs/experience-orchestration.md). After a read-only pre-audit, select and read the actual project design documents and relevant Design.md/design-system references. Run `experience --phase context --surface public-page|app-screen` to stage complete methods and obtain the live audit/decision schemas and evidence references. Record `experience --phase audit --input FILE` before formal diagnosis, then `experience --phase decision --input FILE` after scope confirmation and before `plan`.

Separate the original user request and hard constraints from operator proposals and assumptions. The audit covers the scoped journey, current-page control/AI slop details, content, hierarchy, grid, interactions, recovery and remaining coverage. Keep experience hypotheses separate from evidence-backed defects. The decision must explain reading order, grouping, action, content burden, component relationships and a lighter/preserve alternative. Existing project rules outrank external references; source observation does not establish approval. A reference must identify an actual source hash and section and explain task-specific adoption, preservation and rejection.

The executor may challenge a bad decision before editing. Independent review starts with the original task and images, then checks the operator rationale; it must separately assess task, hierarchy, interaction, design-system and regression. Field completeness, script passes and agreement with an operator's proposal cannot establish improved experience. Keep every unresolved audit problem visible. Retain all previous failures and obtain a real human decision; do not retrofit old runs or automatically make a global aesthetic ban from one rejected example.

