diff --git a/profiles/browser/profile.yaml b/profiles/browser/profile.yaml
index 4f2e86f0..3e96083d 100644
--- a/profiles/browser/profile.yaml
+++ b/profiles/browser/profile.yaml
@@ -5,7 +5,6 @@ inherits: core
 
 skills:
   local:
-    - browser/lightpanda
     - browser/agent-browser   # vercel-labs/agent-browser — Rust browser CLI, ref-based automation, daemon-backed
     - browser/playwright      # drives the Playwright MCP (Chromium/Firefox/WebKit): screenshot-verify UI, repro bugs
   npx:
@@ -16,8 +15,8 @@ skills:
 mcps:
   # Playwright — Microsoft's official Playwright MCP (`npx @playwright/mcp@latest`).
   # Full Chromium/Firefox/WebKit with real screenshots, network log, a11y snapshot;
-  # tools appear as mcp__playwright__*. Pairs with the browser/playwright skill for
-  # pixel-true visual verification (the gap lightpanda can't fill).
+  # tools appear as mcp__playwright__*. Pairs with the browser/playwright skill
+  # when a task needs Firefox or WebKit — ego-browser covers Chromium.
   - playwright
 
 plugins:
@@ -33,18 +32,18 @@ persona: |
   `final_runs/run_<id>/`, and visually self-verify the result before declaring
   the task done.
 
-  ## Default browser engine: Lightpanda
+  ## Default browser engine: ego-browser
 
-  Lightpanda is your default browser. It is a fast, low-memory headless browser
-  exposed over MCP (`mcp__lightpanda__*`) and as a CLI at `/home/deadpool/lightpanda`.
-  Use it for headless navigation, scraping, and DOM/markdown extraction. For pure
-  scrape/render jobs, `lightpanda fetch <url> --dump markdown` is the fastest path.
-  See the `browser/lightpanda` skill for all three modes (mcp / serve / fetch).
+  ego-browser is your default browser, inherited from core. One JS heredoc drives
+  a real Chromium over CDP — navigation, forms, clicks, semantic snapshots with
+  element refs, screenshots, downloads — so a whole interaction costs one bash
+  round trip instead of a dozen tool calls. It reuses your real logins, and each
+  agent gets its own task space. See the `browser/ego-browser` skill.
 
-  Lightpanda does **not** paint pixels. The moment a task needs a real screenshot
-  to visually verify a page, switch to full Chromium via Webwright (below). You can
-  also point Webwright/Playwright at lightpanda as a low-memory CDP backend with
-  `lightpanda serve --port 9222` + `connectOverCDP("ws://127.0.0.1:9222")`.
+  Reach for Webwright (below) when you need its plugin workflow — the run
+  directory, the action log, the pinned Playwright workspace. Reach for the
+  `playwright` MCP when you need Firefox or WebKit specifically; ego-browser is
+  Chromium only.
 
   ## Operating loop
 
diff --git a/profiles/commerce/profile.yaml b/profiles/commerce/profile.yaml
index 40ca388e..7f2586f1 100644
--- a/profiles/commerce/profile.yaml
+++ b/profiles/commerce/profile.yaml
@@ -140,10 +140,11 @@ persona: |
   Three browser engines, distinct sweet spots — pick by the job, don't
   default to one:
 
-  - **lightpanda** (MCP / `lightpanda fetch <url> --dump markdown`) — the
-    lightest. Headless, renders DOM + JS but paints **no pixels**. Use for
-    scraping, DOM/markdown extraction, "what's on this URL", cheap bulk
-    navigation. Fastest, lowest memory. Can't screenshot.
+  - **ego-browser** (JS heredoc) — the default, inherited from core. One bash
+    round trip drives a real Chromium: navigate, fill, click, semantic
+    snapshot with `@ref` handles, screenshot, download. Use for scraping,
+    extraction, "what's on this URL", cheap bulk navigation — and for
+    anything that needs a picture, since it paints pixels too.
   - **agent-browser** (CLI) — the interactive default. `snapshot -i` →
     act on `@ref` → `screenshot`. Use to drive and visually verify one
     flow: login, add-to-cart, checkout QA, "does this storefront render
@@ -153,10 +154,9 @@ persona: |
     artifacts to `final_runs/`. Use for autonomous end-to-end missions,
     not a handful of commands.
 
-  Rule of thumb: scrape/read → lightpanda; drive+verify one flow →
-  agent-browser; autonomous multi-step mission → webwright. (lightpanda
-  can also back agent-browser/webwright as a low-memory CDP engine via
-  `lightpanda serve`.)
+  Rule of thumb: scrape/read or drive+verify one flow → ego-browser; reach
+  for agent-browser when you want its persistent Rust daemon; autonomous
+  multi-step mission → webwright.
 
   ## Surface router
 
diff --git a/profiles/core/profile.yaml b/profiles/core/profile.yaml
index 25ed0860..73a6c374 100644
--- a/profiles/core/profile.yaml
+++ b/profiles/core/profile.yaml
@@ -152,7 +152,6 @@ skills:
     - security/trivy-scan       # /trivy-scan — pre-merge supply-chain gate (Trivy: dependency CVEs + secrets + IaC misconfig). Hard-blocks HIGH/CRITICAL. Invoked by code-review-deep Pass 0 + ship Step 9.0.
     # ── Browser ──
     - browser/ego-browser       # DEFAULT browser path. One JS heredoc drives a real Chromium (navigate, forms, clicks, snapshot, screenshots) instead of many MCP tool-call round trips — chosen for token cost. Reuses the user's real logins. Linux port lives in ~/Documents/ego-lite-linux (package/ego-linux).
-    - browser/lightpanda        # Lightpanda — fast headless browser for scraping, DOM dump, CDP. Pairs with lightpanda MCP.
     # ── Source-fetching ──
     - tools/opensrc             # opensrc — fetch dependency source (npm/PyPI/crates/GitHub) so agents read implementations, not just types
     - tools/context7            # context7 — fetch up-to-date, version-specific library/API docs. Pairs with the context7 MCP (in core's mcps).
diff --git a/profiles/gstack/profile.yaml b/profiles/gstack/profile.yaml
index d2cfdcf3..59b45ccd 100644
--- a/profiles/gstack/profile.yaml
+++ b/profiles/gstack/profile.yaml
@@ -165,7 +165,8 @@ skills:
     # ── GitHub / worktree safety ───────────────────────────────────────────────
     - github/gitguardex
 
-# lightpanda is inherited from core; cue-tty-watch is NOT (core ships lightpanda only).
+# cue-tty-watch is NOT inherited from core — core ships dataforseo, codegraph,
+# context7 and headroom. (It used to ship lightpanda; that MCP is gone.)
 mcps:
   - gbrain   # gstack/setup-gbrain + gstack/sync-gbrain drive the gbrain MCP
 rules:
diff --git a/profiles/webshop/profile.yaml b/profiles/webshop/profile.yaml
index 89d3ee18..d0f3aaeb 100644
--- a/profiles/webshop/profile.yaml
+++ b/profiles/webshop/profile.yaml
@@ -126,7 +126,7 @@ persona: |
   | Storefront UI implementation, visual polish, a11y          | design/* + browser/agent-browser |
   | Premium visual design, brand kit, redesign                 | design/taste-* + design/brandkit |
   | Long-horizon browser automation, e2e flows                 | webwright plugin |
-  | Scrape / headless read, DOM or markdown dump (no pixels)    | lightpanda       |
+  | Scrape / headless read, page snapshot, screenshots          | browser/ego-browser |
   | Product photos, hero shots, marketplace cards, Soul ID     | higgsfield/*     |
   | Transactional + marketing email, React Email, deliverability | resend plugin  |
   | Containers, env vars, reverse proxy, zero-downtime deploy  | deployment/coolify + gstack/setup-deploy |
@@ -136,9 +136,10 @@ persona: |
 
   Three browser engines, distinct sweet spots — pick by the job:
 
-  - **lightpanda** (MCP / `lightpanda fetch <url> --dump markdown`) — the
-    lightest. Headless, renders DOM + JS but paints **no pixels**. Use for
-    scraping, DOM/markdown extraction, cheap bulk navigation. Can't screenshot.
+  - **ego-browser** (JS heredoc) — the default, inherited from core. One bash
+    round trip drives a real Chromium: navigate, fill, click, semantic snapshot
+    with `@ref` handles, screenshot. Use for scraping, extraction, cheap bulk
+    navigation, and anything where a page needs looking at.
   - **agent-browser** (CLI) — the interactive default. `snapshot -i` → act on
     `@ref` → `screenshot`. Use to drive and visually verify one flow: login,
     add-to-cart, checkout QA, "does this storefront render right". Real pixels.
@@ -146,8 +147,9 @@ persona: |
     goal; it plans → acts → screenshot-self-verifies → logs to `final_runs/`.
     Use for autonomous end-to-end missions, not a handful of commands.
 
-  Rule of thumb: scrape/read → lightpanda; drive+verify one flow →
-  agent-browser; autonomous multi-step mission → webwright.
+  Rule of thumb: scrape/read or drive+verify one flow → ego-browser; reach for
+  agent-browser when you want its Rust CLI daemon; autonomous multi-step mission
+  → webwright.
 
   ## Defaults across the stack
 
diff --git a/profiles/x-growth-bot/profile.yaml b/profiles/x-growth-bot/profile.yaml
index 932b5e6a..ee39e1e7 100644
--- a/profiles/x-growth-bot/profile.yaml
+++ b/profiles/x-growth-bot/profile.yaml
@@ -13,7 +13,7 @@ skills:
 
 # The bot's MCP control surface (~27 tools: xbot_status, xbot_think, xbot_reply,
 # xbot_dm, xbot_cycle, xbot_pause, xbot_control, ...). core already contributes
-# cue-tty-watch + lightpanda.
+# codegraph + context7.
 mcps:
   - xbot
 
diff --git a/resources/skills b/resources/skills
index ba3d06db..752aef4d 160000
--- a/resources/skills
+++ b/resources/skills
@@ -1 +1 @@
-Subproject commit ba3d06dbcec1c2cc4fd5b447fd13973565051210
+Subproject commit 752aef4d8b1d1b1db84be61026a222f47c3aa1a5