diff --git a/src/app/api/gigs/route.test.ts b/src/app/api/gigs/route.test.ts
index 3b5bd0e8..c0a6e52c 100644
--- a/src/app/api/gigs/route.test.ts
+++ b/src/app/api/gigs/route.test.ts
@@ -113,6 +113,33 @@ describe("GET /api/gigs", () => {
     expect(json.pagination.total).toBe(1);
   });
 
+  it.each(["human", "agent"])("filters gigs by the %s poster relationship", async (accountType) => {
+    const chain = chainResult({ data: null, error: null });
+    chain.range = vi.fn().mockResolvedValue({ data: [], error: null, count: 0 });
+    mockFrom.mockReturnValue(chain);
+
+    const res = await GET(makeGetRequest({ account_type: accountType }));
+
+    expect(res.status).toBe(200);
+    // Filtering the embedded object alone keeps unrelated parent rows.
+    // An inner join must constrain the gigs and their exact pagination count.
+    expect(chain.select.mock.calls[0][0]).toContain("poster:profiles!poster_id!inner");
+    expect(chain.eq).toHaveBeenCalledWith("poster.account_type", accountType);
+    expect(chain.eq).not.toHaveBeenCalledWith("poster:profiles!poster_id.account_type", accountType);
+  });
+
+  it("keeps the default left join when no account type is requested", async () => {
+    const chain = chainResult({ data: null, error: null });
+    chain.range = vi.fn().mockResolvedValue({ data: [], error: null, count: 0 });
+    mockFrom.mockReturnValue(chain);
+
+    const res = await GET(makeGetRequest());
+
+    expect(res.status).toBe(200);
+    expect(chain.select.mock.calls[0][0]).not.toContain("!inner");
+    expect(chain.eq.mock.calls.some(([column]) => column === "poster.account_type")).toBe(false);
+  });
+
   it("caps huge page values before building the Supabase range", async () => {
     const chain = chainResult({ data: null, error: null });
     chain.select = vi.fn().mockReturnValue(chain);
diff --git a/src/app/api/gigs/route.ts b/src/app/api/gigs/route.ts
index 2beacd8a..44c63a9c 100644
--- a/src/app/api/gigs/route.ts
+++ b/src/app/api/gigs/route.ts
@@ -59,7 +59,7 @@ export async function GET(request: NextRequest) {
       .select(
         `
         *,
-        poster:profiles!poster_id (
+        poster:profiles!poster_id${account_type ? "!inner" : ""} (
           id,
           username,
           full_name,
@@ -121,7 +121,7 @@ export async function GET(request: NextRequest) {
     }
 
     if (account_type) {
-      query = query.eq("poster:profiles!poster_id.account_type", account_type);
+      query = query.eq("poster.account_type", account_type);
     }
 
     // Apply sorting
diff --git a/src/lib/github-links.test.ts b/src/lib/github-links.test.ts
index aa216433..6f6eff3f 100644
--- a/src/lib/github-links.test.ts
+++ b/src/lib/github-links.test.ts
@@ -7,6 +7,13 @@ import {
 } from "./github-links";
 
 describe("isGitHubPrLink", () => {
+  it("rejects a numeric prefix followed by a non-path suffix", () => {
+    for (const suffix of ["12abc", "12.5", "12-closed"]) {
+      const url = `https://github.com/org/repo/pull/${suffix}`;
+      expect(isGitHubPrLink(url)).toBe(false);
+      expect(parseGitHubPullUrl(url)).toBeNull();
+    }
+  });
   it("accepts a single pull request URL", () => {
     expect(isGitHubPrLink("https://github.com/profullstack/ugig.net/pull/42")).toBe(true);
     expect(isGitHubPrLink("https://github.com/org/repo/pull/1/files")).toBe(true);
diff --git a/src/lib/github-links.ts b/src/lib/github-links.ts
index fa20a75a..8898398e 100644
--- a/src/lib/github-links.ts
+++ b/src/lib/github-links.ts
@@ -14,7 +14,7 @@ export function isGitHubPrLink(value: string): boolean {
   const path = url.pathname;
   return (
     // A single PR: /owner/repo/pull/123 (optionally /files, #discussion, …)
-    /^\/[^/]+\/[^/]+\/pull\/\d+/.test(path) ||
+    /^\/[^/]+\/[^/]+\/pull\/\d+(?:\/.*)?$/.test(path) ||
     // A repo's PR list/search: /owner/repo/pulls
     /^\/[^/]+\/[^/]+\/pulls\/?$/.test(path) ||
     // The global PR search: /pulls
diff --git a/src/lib/mentions.test.ts b/src/lib/mentions.test.ts
index 373ae9f4..8aab8948 100644
--- a/src/lib/mentions.test.ts
+++ b/src/lib/mentions.test.ts
@@ -35,6 +35,21 @@ describe("parseMentions", () => {
 });
 
 describe("parseContentWithMentions", () => {
+  it("keeps email addresses and URL userinfo as ordinary text", () => {
+    for (const content of ["Contact alice@example.com", "https://user@example.com/path"]) {
+      expect(parseContentWithMentions(content)).toEqual([{ type: "text", value: content }]);
+      expect(parseMentions(content)).toEqual([]);
+    }
+  });
+
+  it("preserves parentheses and whitespace around real mentions", () => {
+    expect(parseContentWithMentions("Hi (@Alice) and\n@bob")).toEqual([
+      { type: "text", value: "Hi (" },
+      { type: "mention", username: "Alice" },
+      { type: "text", value: ") and\n" },
+      { type: "mention", username: "bob" },
+    ]);
+  });
   it("parses text with mentions into segments", () => {
     const segments = parseContentWithMentions("hey @alice check this");
     expect(segments).toEqual([
diff --git a/src/lib/mentions.ts b/src/lib/mentions.ts
index 8e2fb87f..47133c76 100644
--- a/src/lib/mentions.ts
+++ b/src/lib/mentions.ts
@@ -24,15 +24,18 @@ export type MentionSegment =
 
 export function parseContentWithMentions(content: string): MentionSegment[] {
   const segments: MentionSegment[] = [];
-  const regex = /@([a-zA-Z0-9_-]+)/g;
+  // Use the same token boundary as notification extraction. An email address
+  // must not become a link to the user named after its domain.
+  const regex = /(^|[\s(])@([a-zA-Z0-9_-]+)/g;
   let lastIndex = 0;
   let match: RegExpExecArray | null;
 
   while ((match = regex.exec(content)) !== null) {
-    if (match.index > lastIndex) {
-      segments.push({ type: "text", value: content.slice(lastIndex, match.index) });
+    const mentionStart = match.index + match[1].length;
+    if (mentionStart > lastIndex) {
+      segments.push({ type: "text", value: content.slice(lastIndex, mentionStart) });
     }
-    segments.push({ type: "mention", username: match[1] });
+    segments.push({ type: "mention", username: match[2] });
     lastIndex = regex.lastIndex;
   }