diff --git a/SKILL.md b/SKILL.md
index 5255812..efd5073 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -376,6 +376,27 @@ Request Id. Full field mapping, prerequisites, and the exact browser steps are i
 Because the files upload from local disk, commit each component's `.SchLib`/`.PcbLib`/datasheet
 to the operator's machine (device bridge) first, and put those local paths in the manifest.
 
+### Fully automatic end-of-run (no operator action)
+
+To make the Part Request the automatic tail of every run, the end task doesn't drive the browser
+from the cloud (it can't reach the operator's desktop Chrome). Instead:
+
+1. After the Gitea push, the skill **writes `part_requests.json`** for the component(s) processed
+   this run, and **commits it plus each component's `.SchLib`/`.PcbLib`/datasheet to an *inbox
+   folder on the operator's machine*** via the device bridge (e.g.
+   `.../altium-library-master/_part_request_inbox/`). Use the local file paths (under that inbox)
+   in the manifest's `files`.
+2. A one-time **local runner** on the operator's Windows machine (`scripts/local_runner.py`, set
+   up as a Task Scheduler job at logon or a startup shortcut, configured via
+   `runner_config.json`) watches that inbox, ensures Chrome is up with the debug port on the
+   signed-in profile, runs `altium365_part_request.py` on the manifest, and archives it.
+
+So each run ends with the skill dropping the manifest+files into the inbox; the local runner
+submits the Part Request on its own. The only standing requirements are that the operator's
+machine is on and its Chrome profile has been signed into the Workspace at least once (the runner
+relaunches Chrome with the debug port and the persisted session is reused). This keeps the
+per-component Claude-token cost at zero for the submission step.
+
 ## Per-typeid versioning
 
 Versioning is **per typeid**, not global. Each typeid carries its own `template_version` and
@@ -542,6 +563,10 @@ plain flat push, but it does not merge the changelog or blank the token, so pref
 - `scripts/altium365_part_request.py` — standalone **Selenium** submitter: attaches to the
   operator's signed-in Chrome and loops the `part_requests.json` manifest, filling and saving each
   Part Request. Token-free per component; the token-economical end task for a whole library.
+- `scripts/local_runner.py` (+ `runner_config.example.json`) — the operator installs this **once**
+  on their machine; it watches the inbox for manifests the skill drops there, ensures Chrome's
+  debug session, runs the Selenium submitter, and archives each manifest — making submission the
+  automatic, hands-off tail of every run.
 - `assets/template/versions.json` — per-typeid `template_version` + `skill_version`.
 - `assets/CHANGELOG.xlsx` — global version/parameter changelog (created on first add;
   merged into the skill repo's copy in Gitea by `push-skill`).
diff --git a/assets/CHANGELOG.xlsx b/assets/CHANGELOG.xlsx
index 39139a0..ad4994c 100644
Binary files a/assets/CHANGELOG.xlsx and b/assets/CHANGELOG.xlsx differ
diff --git a/scripts/local_runner.py b/scripts/local_runner.py
new file mode 100644
index 0000000..7020839
--- /dev/null
+++ b/scripts/local_runner.py
@@ -0,0 +1,104 @@
+#!/usr/bin/env python3
+"""LOCAL RUNNER — runs on the operator's own machine (Windows), set up ONCE, to make Altium 365
+Part Request submission the automatic tail of a skill run.
+
+Why this exists: the library-manager skill runs in the cloud, but the Part Request web form must
+be driven from the operator's signed-in Chrome on their desktop — the cloud can't reach it. So at
+the end of a run the skill drops the finished files + a `part_requests.json` manifest into an
+**inbox folder on this machine** (via the device bridge). This runner watches that inbox and, for
+each manifest: makes sure Chrome is running with a debugging port on the operator's profile
+(launching it if needed — the persisted login is reused), runs `altium365_part_request.py`, then
+moves the manifest to a `processed/` folder.
+
+Set it up once (Task Scheduler at logon, or a startup shortcut) and every future skill run
+auto-submits its Part Requests with no further action — as long as this machine is on and the
+Chrome profile has been signed into the Workspace at least once.
+
+Config: a `runner_config.json` next to this file, e.g.
+{
+  "base": "https://vecmocon-technologies-pvt-ltd.365.altium.com",
+  "inbox": "D:/User Data/Desktop/altium-library-master/_part_request_inbox",
+  "processed": "D:/User Data/Desktop/altium-library-master/_part_request_inbox/processed",
+  "chrome": "C:/Program Files/Google/Chrome/Application/chrome.exe",
+  "user_data_dir": "%LOCALAPPDATA%/Google/Chrome/User Data",
+  "debug_port": 9222,
+  "chromedriver": null,
+  "review_first": false,
+  "poll_seconds": 30
+}
+
+Run:
+  python local_runner.py           # watch the inbox forever (use for Task Scheduler / startup)
+  python local_runner.py --once    # process whatever is in the inbox now, then exit
+"""
+import glob, json, os, shutil, socket, subprocess, sys, time
+
+HERE = os.path.dirname(os.path.abspath(__file__))
+
+
+def load_cfg():
+    with open(os.path.join(HERE, "runner_config.json"), encoding="utf-8") as f:
+        return json.load(f)
+
+
+def _port_open(port):
+    s = socket.socket(); s.settimeout(1)
+    try:
+        s.connect(("127.0.0.1", int(port))); return True
+    except Exception:
+        return False
+    finally:
+        s.close()
+
+
+def ensure_chrome(cfg):
+    """Make sure a Chrome with the debug port is up (reusing the operator's logged-in profile)."""
+    if _port_open(cfg["debug_port"]):
+        return
+    udd = os.path.expandvars(cfg["user_data_dir"])
+    subprocess.Popen([cfg["chrome"], f"--remote-debugging-port={cfg['debug_port']}",
+                      f"--user-data-dir={udd}"])
+    for _ in range(40):
+        if _port_open(cfg["debug_port"]):
+            time.sleep(2)                       # give the profile a moment to restore the session
+            return
+        time.sleep(1)
+    raise RuntimeError("Chrome did not expose a debug port — check 'chrome' path and 'debug_port'.")
+
+
+def process(cfg, manifest):
+    ensure_chrome(cfg)
+    cmd = [sys.executable, os.path.join(HERE, "altium365_part_request.py"),
+           "--manifest", manifest, "--base", cfg["base"],
+           "--debugger-address", f"127.0.0.1:{cfg['debug_port']}",
+           "--out", manifest + ".result.json"]
+    if cfg.get("chromedriver"):
+        cmd += ["--chromedriver", cfg["chromedriver"]]
+    if cfg.get("review_first"):
+        cmd += ["--review-first"]
+    subprocess.run(cmd, check=False)
+
+
+def main():
+    cfg = load_cfg()
+    os.makedirs(cfg["inbox"], exist_ok=True)
+    os.makedirs(cfg["processed"], exist_ok=True)
+    once = "--once" in sys.argv
+    print(f"watching {cfg['inbox']} (base {cfg['base']})")
+    while True:
+        for m in sorted(glob.glob(os.path.join(cfg["inbox"], "*.json"))):
+            if m.endswith(".result.json"):
+                continue
+            print("processing", m)
+            try:
+                process(cfg, m)
+            except Exception as e:
+                print("  error:", e)
+            shutil.move(m, os.path.join(cfg["processed"], os.path.basename(m)))
+        if once:
+            break
+        time.sleep(int(cfg.get("poll_seconds", 30)))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/runner_config.example.json b/scripts/runner_config.example.json
new file mode 100644
index 0000000..6f79490
--- /dev/null
+++ b/scripts/runner_config.example.json
@@ -0,0 +1,11 @@
+{
+  "base": "https://vecmocon-technologies-pvt-ltd.365.altium.com",
+  "inbox": "D:/User Data/Desktop/altium-library-master/_part_request_inbox",
+  "processed": "D:/User Data/Desktop/altium-library-master/_part_request_inbox/processed",
+  "chrome": "C:/Program Files/Google/Chrome/Application/chrome.exe",
+  "user_data_dir": "%LOCALAPPDATA%/Google/Chrome/User Data",
+  "debug_port": 9222,
+  "chromedriver": null,
+  "review_first": false,
+  "poll_seconds": 30
+}