New to Claude Skills? Learn how to install them →

forcedotcom on GitHub

LWC RTL Review

OfficialFree

Ensure your Lightning Web Components are RTL compliant.

Get this skill

Free · Opens the source repo

What LWC RTL Review does

The LWC RTL Review skill is designed to help developers ensure that their Lightning Web Components (LWCs) are compliant with right-to-left (RTL) internationalization standards. This skill performs a thorough review of LWC files, including HTML, JavaScript, and CSS, to identify potential issues related to RTL layouts, bidirectional text handling, and the correct usage of Salesforce Lightning Design System (SLDS) classes. By automating this review process, developers can save time and reduce the risk of errors when preparing components for RTL locales such as Arabic and Hebrew.

When you trigger the skill by asking for an RTL review or an i18n compliance check, it begins by scoping the review based on the component path provided. The skill inspects the relevant files, applying a structured approach to identify CSS properties that need to be converted to logical properties, ensuring that the layout behaves correctly in RTL contexts. It also checks for proper handling of directional text and icons, ensuring that user interfaces are intuitive and accessible for RTL users.

The workflow includes running a dedicated CSS scanner that identifies physical-to-logical property mappings and flags any inline styles that may not comply with RTL standards. Additionally, the skill verifies that SLDS class usage aligns with Salesforce’s guidelines, ensuring that developers do not inadvertently modify classes that are already RTL-aware. Overall, this skill is an essential tool for developers looking to enhance the internationalization of their LWCs and provide a seamless experience for users in RTL locales.

When to use it

Use this skill when preparing an LWC for release in RTL locales or when investigating reported layout defects in RTL environments.

When not to use it

This skill is not suitable for building new components or modifying SLDS classes directly; use the appropriate skills for those tasks instead.

What you can build with it

Preparing for RTL Release

Use this skill to audit your LWC before releasing it in RTL locales like Arabic or Hebrew.

Investigating Layout Issues

When users report layout defects in RTL environments, run this skill to identify and fix the issues.

Verifying SLDS Class Usage

After modifying CSS, check SLDS class usage to ensure compliance with Salesforce's RTL guidelines.

How to install LWC RTL Review

View source

1. Install with the skills CLI

npx skills add forcedotcom/sf-skills/experience-lwc-rtl-validate --agent claude-code

2. Or install it manually

Download the skill folder and drop it into ~/.claude/skills/ for all projects, or .claude/skills/ to scope it to one repo. Restart Claude Code so it picks up the new skill.

Anthropic's agentic coding CLI, and the reference implementation of Agent Skills. Drop a skill folder into ~/.claude/skills and Claude Code loads it automatically whenever a task matches the skill's description. Claude Code docs

Inside SKILL.md

Written by forcedotcom
<!-- adk-managed-skill -->

Reviewing LWC RTL

Run a structured right-to-left (RTL) internationalization compliance pass over a Lightning Web Component, producing a report of issues found and code-level fixes to bring the component into compliance with Salesforce RTL guidelines.

When to Use

  • The user asks for an "RTL review", "i18n check", "RTL compliance pass", or "RTL audit" on a specific LWC.
  • Preparing a component for release in RTL locales (Arabic, Hebrew, Farsi, Urdu).
  • Investigating layout defects reported in RTL environments.
  • Verifying SLDS class usage after a CSS refactor.

Do NOT use this skill for:

  • Building new components (use experience-lwc-generate).
  • Modifying SLDS classes themselves (use design-systems-slds-apply or design-systems-slds2-migrate).
  • Accessibility or security review — run those as separate passes with the relevant tooling.
  • Gating a fix behind a feature flag (apply feature-flag gating after fixes land).

Prerequisites

  • Component path (LWC bundle under modules/…).
  • Access to the component's HTML templates, JS/TS, and CSS.

Knowledge Base

The reference is the source of truth. Do not summarize from memory — open the reference, apply the guidelines, and cite the specific section you used in the report.

Workflow

Step 1 — Scope the review

Collect the component path and identify the files to review: .html, .js/.ts, .css, and any child components owned by the same team that are invoked from the target.

Note any existing feature-flag gates (e.g., Aura.org.rtlPhase1FixEnabled) — findings that require code changes must respect them.

Step 2 — Read the knowledge base

Read RTL Expert top-to-bottom before judging. It enumerates the physical-to-logical property mappings, bidirectional text handling patterns, and — critically — the SLDS constraints that override generic RTL advice.

Step 3 — CSS inspection

Run the deterministic scanner over every .css file in the bundle. The scanner matches CSS declarations only — never SLDS class names in HTML class="…" attributes (see Step 5). Inline style="…" attributes must be scanned separately (either by extracting them into a temp file or by inspecting the HTML by hand and applying the same rules).

"<skill_dir>/scripts/scan-rtl-css.sh" <cssFile1> [<cssFile2> ...]

Each output line has the shape <file>:<line>: <property>: <value> -> <logical-property>: <logical-value>. The scanner tokenizes declarations (splits on ; inside {…}) so minified multi-declaration lines produce one finding per physical declaration and selector names are never rewritten. Translate each line into a Step 6 bullet (<file>:<line> — <pattern> → <logical property> plus a one-sentence Fix). Empty output IS a valid result — record it as "No issues found." in the report.

Scanner-recognised patterns (kept in sync with the script's regex — do not add rules here without also updating scan-rtl-css.sh):

  • left / rightinset-inline-start / inset-inline-end
  • margin-left / margin-rightmargin-inline-start / margin-inline-end
  • padding-left / padding-rightpadding-inline-start / padding-inline-end
  • text-align: left / righttext-align: start / end
  • border-left-* / border-right-*border-inline-start-* / border-inline-end-*
  • float: left / float: rightfloat: inline-start / float: inline-end (or remove and use flex/grid)
  • transform: translateX(...) — flagged; sign-flip or use a logical alternative

border-radius with explicit corners is not scanned automatically — inspect corner shorthand by hand and translate to the logical corner variants.

Step 4 — HTML / JS inspection

Walk templates and JS for:

  • Icon flipping hints — flag only icons with directional semantics that appear in plain HTML <img> / inline SVG / raw Unicode / background-image CSS. Do NOT flag <lightning-icon> in LWC templates (including directional utility names like utility:chevronright, utility:chevronleft, utility:back, utility:forward); lightning-icon renders through the Lightning icon service, which mirrors directional utility icons automatically in RTL locales. Treat it the same way you treat SLDS utility classes.
  • dir attribute usage — confirm it's sourced from locale, not hardcoded.
  • Keyboard arrow-key semantics — Left/Right arrow handlers should swap in RTL where navigation is directional (tab bars, sliders, tree expand/collapse).
  • Directional Unicode controls — ensure user-generated text is not stripped of RLM/LRM markers when rendered.
  • Inline style="…" with physical properties — same rules as Step 3.

Step 5 — SLDS constraints

SLDS class handling is the highest-priority RTL rule. When a rule below says "do NOT flag", that class must not appear as a finding in the report — not as an issue, not as a "fix", not as a rename suggestion.

  1. SLDS utility classes with _left / _right suffixes are RTL-aware and MUST NOT be flagged. They already mirror automatically in RTL under the hood — this includes slds-text-align_right, slds-text-align_left, slds-m-left_*, slds-m-right_*, slds-p-left_*, slds-p-right_*, slds-float_left, slds-float_right, slds-border_left, slds-border_right, and the rest of the _left / _right utility family. Leave them exactly as written.
  2. _start / _end variants of SLDS classes DO NOT EXIST. Never rename slds-*_rightslds-*_end or slds-*_leftslds-*_start. There is no such class; the rename would break the stylesheet. If you are tempted to "fix" an SLDS utility class by adding _start/_end, STOP — the correct action is to leave the class alone (see rule 1).
  3. Never modify, rename, or remove any slds-* class. SLDS ships RTL-aware stylesheets; altering classes breaks the contract.
  4. If custom CSS duplicates what an SLDS class already handles, the fix is removal — not conversion. Delete the redundant custom rule; leave the SLDS class alone. Do NOT convert the removed property to its logical equivalent, and do NOT offer removal + conversion as two alternatives; there is one fix, and it is deletion. Example: float: right in a .css file next to class="slds-button__icon_right" in the template — delete the entire .custom-icon { float: right } rule. Do not "also suggest float: inline-end" — that would be wrong; the SLDS class already handles placement.
  5. If an SLDS class has an RTL gap, add complementary custom CSS rather than altering the SLDS class.

Concrete negative example — a template with <span class="slds-text-align_right slds-m-right_small">…</span> alongside a custom padding-left: 4px in the component's CSS: the ONLY finding is the CSS padding-left (Step 3). The two slds-*_right classes are not findings and must not appear in the report.

Step 6 — Produce the report

Write <outputDir>/rtl-review.md in this exact shape (blank line after each ## heading; every summary sentence ends with a period):

## RTL

- <file>:<line> — <physical pattern> → <logical property / SLDS class>
  Fix: <one-sentence explanation>; applied: yes/no
- <file>:<line> — <physical pattern> → <logical property / SLDS class>
  Fix: <one-sentence explanation>; applied: yes/no

## Summary

- <n> issues found; <m> fixed; <k> deferred (with reason).
- <one-sentence overview of what was found and why the remaining items are correct or deferred>.
- Cite: RTL Expert — <section title>.

Populate the ## RTL bullets from the scan-rtl-css.sh output for CSS findings, and append hand-authored bullets in the same shape for the Step 4 HTML/JS findings. When there are no findings, emit the single bullet - No issues found. under ## RTL and a Summary that reports 0 issues found; 0 fixed; 0 deferred. plus a one-sentence overview and citation.

Step 7 — Apply fixes

For each accepted finding:

  1. Edit the component files (CSS, HTML, JS/TS) to apply the fix.
  2. Preserve existing correct behavior and existing feature-flag gates. If a gate is already configured (e.g., Aura.org.rtlPhase1FixEnabled), leave it untouched. Only when the caller has explicitly required a phased rollout for a new fix, wrap that fix behind a feature flag; otherwise apply the fix directly.
  3. Do NOT remove or rename SLDS classes.
  4. Do NOT silently delete old code — preserve the original path where a gate is required.

Step 8 — Verify

  • Re-run the review against the updated files; every fixed finding must no longer appear.
  • Run Jest tests and any component-level RTL visual tests.

Cross-References

  • Related skills:
    • design-systems-slds-apply — for SLDS class-level changes beyond custom CSS cleanup.
    • design-systems-slds2-migrate — when RTL cleanup surfaces classes that need SLDS2 migration first.
    • experience-lwc-generate — when the review surfaces a need to regenerate the component rather than patch it.
    • Accessibility (WCAG 2.2) and security (LWS + Product Security) reviews are separate passes — run them with the appropriate tooling for each, not this skill.
    • Feature-flag gating is out of scope for this skill; apply it separately only when the caller has explicitly required a phased rollout (see Step 7).

Verification

  • All RTL rules are green on the updated component (re-run scripts/scan-rtl-css.sh against every CSS file in the bundle — it must return empty for every fully-applied finding).
  • No SLDS class was modified, renamed, or removed.
  • Every finding has either been applied or has an explicit deferred note with a reason.
  • The rtl-review.md matches the two-section shape defined in Step 6 (blank line after each ##; period-terminated summary sentences).

Frequently asked questions about LWC RTL Review

Similar skills