Merge branch 'main' into twishabansal-patch-3

Author: averikitsch

.github/workflows/markdown-checks.yml

@@ -27,7 +27,7 @@ jobs:
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5
 
       - name: Link Checker
-        uses: lycheeverse/lychee-action@885c65f3dc543b57c898c8099f4e08c8afd178a2 # v2.6.1
+        uses: lycheeverse/lychee-action@a8c4c7cb88f0c7386610c35eb25108e448569cb0 # v2.7.0
         with:
           # There is no security token. So, it would fail on any links which aren't public.
           args: "--verbose --no-progress **/*.md"

.github/workflows/mirror-changelog.yml

@@ -82,7 +82,25 @@ jobs:
               // Match and extract changelog item
               const itemMatch = trimmedLine.match(/^[*-]\s(.*)$/);
               if (itemMatch) {
-                const originalContent = itemMatch[1]; 
+                let originalContent = itemMatch[1]; 
+
+                // This is a two-step process to prevent `release-please` from
+                // creating mangled, doubly-nested links for PRs.
+
+                // STEP 1: CLEAN THE INPUT.
+                // The source changelog contains a zero-width space as an HTML entity (`​`).
+                // This breaks the regex in the next step, so we must remove it first.
+                // E.g., "[#​1770](...)" becomes "[#1770](...)"
+                originalContent = originalContent.replace(/​/g, '');
+
+                // STEP 2: PROTECT THE OUTPUT.
+                // `release-please` aggressively tries to auto-link any text that looks like `#1234`.
+                // To prevent this, we insert an invisible Unicode zero-width space (`\u200B`)
+                // between the '#' and the number in the link text. This breaks the parser's
+                // pattern matching without changing the visual appearance of the link.
+                // E.g., "[#1770](...)" becomes "[genai-toolbox#​1770](...)"
+                originalContent = originalContent.replace(/\[#(\d+)\](\([^)]+\))/g, '[genai-toolbox#\u200B$1]$2');
+                
                 const lineAsLowerCase = originalContent.toLowerCase();
 
                 const hasPrefix = prefixesToFilter.some(prefix => lineAsLowerCase.includes(prefix));