feat: internationalize AI chat panel strings

Extract ~42 hardcoded English strings from AIChatPanel.js into
src/nls/root/strings.js using AI_CHAT_* keys. Replace all in-code
literals with Strings.* constants and StringUtils.format() for
parameterized text.

Fix a locale-sensitive bug where the Undo button click handler
compared $(this).text() against English "Undo" — now uses
data-action attributes instead.

Pass brackets.getLocale() to the Claude agent so non-English users
get responses in their display language.

Add i18n guidance section to CLAUDE.md.

Author: abose

CLAUDE.md

@@ -13,6 +13,14 @@
 - No trailing whitespace.
 - Use `const` and `let` instead of `var`.
 
+## Translations / i18n
+- All user-visible strings must go in `src/nls/root/strings.js` — never hardcode English in source files.
+- Use `const Strings = require("strings");` then `Strings.KEY_NAME`.
+- For parameterized strings use `StringUtils.format(Strings.KEY, arg0, arg1)` with `{0}`, `{1}` placeholders.
+- Keys use UPPER_SNAKE_CASE grouped by feature prefix (e.g. `AI_CHAT_*`).
+- Only `src/nls/root/strings.js` (English) needs manual edits — other locales are auto-translated by GitHub Actions.
+- Never compare `$(el).text()` against English strings for logic — use data attributes or CSS classes instead.
+
 ## Phoenix MCP (Desktop App Testing)
 
 Use `exec_js` to run JS in the Phoenix browser runtime. jQuery `$()` is global. `brackets.test.*` exposes internal modules (DocumentManager, CommandManager, ProjectManager, FileSystem, EditorManager). Always `return` a value from `exec_js` to see results. Prefer reusing an already-running Phoenix instance (`get_phoenix_status`) over launching a new one.

src-node/claude-code-agent.js

@@ -125,7 +125,7 @@ exports.checkAvailability = async function () {
  *   aiProgress, aiTextStream, aiToolEdit, aiError, aiComplete
  */
 exports.sendPrompt = async function (params) {
-    const { prompt, projectPath, sessionAction, model } = params;
+    const { prompt, projectPath, sessionAction, model, locale } = params;
     const requestId = Date.now().toString(36) + Math.random().toString(36).slice(2, 7);
 
     // Handle session
@@ -142,7 +142,7 @@ exports.sendPrompt = async function (params) {
     currentAbortController = new AbortController();
 
     // Run the query asynchronously — don't await here so we return requestId immediately
-    _runQuery(requestId, prompt, projectPath, model, currentAbortController.signal)
+    _runQuery(requestId, prompt, projectPath, model, currentAbortController.signal, locale)
         .catch(err => {
             console.error("[Phoenix AI] Query error:", err);
         });
@@ -176,7 +176,7 @@ exports.destroySession = async function () {
 /**
  * Internal: run a Claude SDK query and stream results back to the browser.
  */
-async function _runQuery(requestId, prompt, projectPath, model, signal) {
+async function _runQuery(requestId, prompt, projectPath, model, signal, locale) {
     let editCount = 0;
     let toolCounter = 0;
     let queryFn;
@@ -209,7 +209,11 @@ async function _runQuery(requestId, prompt, projectPath, model, signal) {
             "to create brand new files that do not exist yet. For existing files, always use " +
             "multiple Edit calls to make targeted changes rather than rewriting the entire " +
             "file with Write. This is critical because Write replaces the entire file content " +
-            "which is slow and loses undo history.",
+            "which is slow and loses undo history." +
+            (locale && !locale.startsWith("en")
+                ? "\n\nThe user's display language is " + locale + ". " +
+                  "Respond in this language unless they write in a different language."
+                : ""),
         includePartialMessages: true,
         abortController: currentAbortController,
         hooks: {