diff --git a/AGENTS.md b/AGENTS.md
index d27811370fd..09ed6423ac4 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -102,20 +102,6 @@
 - Pure test additions/fixes generally do **not** need a changelog entry unless they alter user-facing behavior or the user asks for one.
 - Mobile: before using a simulator, check for connected real devices (iOS + Android) and prefer them when available.
 
-### Local full-CI one-shot record (2026-02-25)
-
-- Ran local parity sweep for GitHub CI workflows (`workflow-sanity.yml`, `ci.yml`, `install-smoke.yml`, `sandbox-common-smoke.yml`) via `/tmp/run-openclaw-local-ci.sh`.
-- Artifacts: `/tmp/openclaw-local-ci-20260225-221535.summary`, `/tmp/openclaw-local-ci-20260225-221535.log`.
-- Result: `PASS=15`, `FAIL=15`, `SKIP=1` (`ci:secrets-zizmor` skipped because no workflow-file diff vs `origin/main`).
-- `ci.yml` jobs with `if: false` (`deadcode`, `ios`) are disabled in GitHub CI and were not executed.
-- Main local blockers seen in this run:
-  - `ci:check`: formatting failure in `src/daemon/launchd.ts`.
-  - `ci:checks-node-test` and `ci:macos-ts-tests`: test failures plus Node OOM during full `pnpm test`.
-  - `ci:skills-python-*` / `ci:secrets-*`: Homebrew Python PEP668 (`externally-managed-environment`) and missing `pre-commit`/`detect-secrets`.
-  - `ci:macos-swift*`: missing `swiftlint`/`swiftformat` and Swift `6.2.0` required vs local `6.1.0`.
-  - `ci:android-*`: Android SDK not configured (`ANDROID_HOME` / `apps/android/local.properties`).
-  - `sandbox-common-smoke:common-image`: passed after running the workflow command directly (temp script had unescaped `$u` with `set -u`).
-
 ## Commit & Pull Request Guidelines
 
 **Full maintainer PR workflow (optional):** If you want the repo's end-to-end maintainer workflow (triage order, quality bar, rebase rules, commit/changelog conventions, co-contributor policy, and the `review-pr` > `prepare-pr` > `merge-pr` pipeline), see `.agents/skills/PR_WORKFLOW.md`. Maintainers may use other workflows; when a maintainer specifies a workflow, follow that. If no workflow is specified, default to PR_WORKFLOW.
diff --git a/CHANGELOG.md b/CHANGELOG.md
index b890896f0d3..21e75f9ab6d 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -10,6 +10,7 @@ Docs: https://docs.openclaw.ai
 
 ### Fixes
 
+- Daemon/macOS launchd: forward proxy env vars into supervised service environments, switch LaunchAgent keepalive policy to crash-only with throttling, and harden restart sequencing to `print -> bootout -> wait old pid exit -> bootstrap -> kickstart`. (#27276) thanks @frankekn.
 - Android/Node invoke: remove native gateway WebSocket `Origin` header to avoid false origin rejections, unify invoke command registry/policy/error parsing paths, and keep command availability checks centralized to reduce dispatcher/advertisement drift. (#27257) Thanks @obviyus.
 - CI/Windows: shard the Windows `checks-windows` test lane into two matrix jobs and honor explicit shard index overrides in `scripts/test-parallel.mjs` to reduce CI critical-path wall time. (#27234) Thanks @joshavant.
 
diff --git a/src/daemon/launchd.test.ts b/src/daemon/launchd.test.ts
index d302a15de64..7465666a158 100644
--- a/src/daemon/launchd.test.ts
+++ b/src/daemon/launchd.test.ts
@@ -12,6 +12,7 @@ import {
 const state = vi.hoisted(() => ({
   launchctlCalls: [] as string[][],
   listOutput: "",
+  printOutput: "",
   bootstrapError: "",
   dirs: new Set<string>(),
   files: new Map<string, string>(),
@@ -36,6 +37,9 @@ vi.mock("./exec-file.js", () => ({
     if (call[0] === "list") {
       return { stdout: state.listOutput, stderr: "", code: 0 };
     }
+    if (call[0] === "print") {
+      return { stdout: state.printOutput, stderr: "", code: 0 };
+    }
     if (call[0] === "bootstrap" && state.bootstrapError) {
       return { stdout: "", stderr: state.bootstrapError, code: 1 };
     }
@@ -72,6 +76,7 @@ vi.mock("node:fs/promises", async (importOriginal) => {
 beforeEach(() => {
   state.launchctlCalls.length = 0;
   state.listOutput = "";
+  state.printOutput = "";
   state.bootstrapError = "";
   state.dirs.clear();
   state.files.clear();
@@ -224,6 +229,42 @@ describe("launchd install", () => {
     expect(bootstrapIndex).toBeLessThan(kickstartIndex);
   });
 
+  it("waits for previous launchd pid to exit before bootstrapping", async () => {
+    const env = createDefaultLaunchdEnv();
+    state.printOutput = ["state = running", "pid = 4242"].join("\n");
+    const killSpy = vi.spyOn(process, "kill");
+    killSpy
+      .mockImplementationOnce(() => true)
+      .mockImplementationOnce(() => {
+        const err = new Error("no such process") as NodeJS.ErrnoException;
+        err.code = "ESRCH";
+        throw err;
+      });
+
+    vi.useFakeTimers();
+    try {
+      const restartPromise = restartLaunchAgent({
+        env,
+        stdout: new PassThrough(),
+      });
+      await vi.advanceTimersByTimeAsync(250);
+      await restartPromise;
+      expect(killSpy).toHaveBeenCalledWith(4242, 0);
+      const domain = typeof process.getuid === "function" ? `gui/${process.getuid()}` : "gui/501";
+      const label = "ai.openclaw.gateway";
+      const bootoutIndex = state.launchctlCalls.findIndex(
+        (c) => c[0] === "bootout" && c[1] === `${domain}/${label}`,
+      );
+      const bootstrapIndex = state.launchctlCalls.findIndex((c) => c[0] === "bootstrap");
+      expect(bootoutIndex).toBeGreaterThanOrEqual(0);
+      expect(bootstrapIndex).toBeGreaterThanOrEqual(0);
+      expect(bootoutIndex).toBeLessThan(bootstrapIndex);
+    } finally {
+      vi.useRealTimers();
+      killSpy.mockRestore();
+    }
+  });
+
   it("shows actionable guidance when launchctl gui domain does not support bootstrap", async () => {
     state.bootstrapError = "Bootstrap failed: 125: Domain does not support specified action";
     const env = createDefaultLaunchdEnv();