diff --git a/CHANGELOG.md b/CHANGELOG.md
index c6adbb6..324e4fa 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -17,6 +17,7 @@
 
 ### 新增
 
+- 新增平台「广东公需课」（`platforms/gdgxk.py`）：`jsglpt.gds.edu.cn` 账号密码（base64）登录后换 ITLMS 票进 `jsxx.gds.edu.cn`，课程列表读 `/uc/store/courseRegister`；学习只跑 `updateViewTime1` 进度心跳（服务端按 `lastViewTime` 直接记账，不需要真播）。配套 `script/capture-jsglpt.py`（抓包取证）与 `script/test-gdgxk.py`（接口自测）
 - 统一日志源 `runtime.log()`：控制台、内存（`/api/log`）、SQLite（`/api/log/history`）同一份内容；新增右上角「系统日志」弹窗查看 `key=system` 的启动/设置类日志
 - 设置保存、服务启动都写入统一日志；卡片「课程日志」和「系统日志」的「清空」会同时清内存和 SQLite
 - 前端：卡片状态徽标、学习进度、课程勾选、卡片内实时日志、测验 / 设置 / 系统日志弹窗全部保留；新增平台名徽标与空态引导
diff --git a/OPTIMIZATION.md b/OPTIMIZATION.md
index 01bff25..285e751 100644
--- a/OPTIMIZATION.md
+++ b/OPTIMIZATION.md
@@ -171,3 +171,39 @@ GET /mod/school/wanpeng/syncSelf?courseCode=<courseCode>
 
 - 控制台没有 `[system]` 行 → 跑的是旧进程（端口被占用时新代码会提示后直接退出），先结束旧 PID 再启动。
 - 「系统日志」弹窗空白 → `GET /api/log?key=system` 有 lines 就是前端缓存问题，刷新即可；没有 lines 说明还没产生 system 日志，保存一次设置就能看到。
+
+## 广东公需课平台接入（jsglpt → jsxx）
+
+**状态**：接口链路已实测跑通（课程列表、播放注册、进度心跳都拿到 `responseCode=00`），待用真实密码登录 + 未学完课程确认端到端。
+
+### 已确认的接口
+
+- 登录：`POST https://jsglpt.gds.edu.cn/loginIn.do`，`loginUser.userName` / `loginUser.password` 都是 **base64**（账号是手机号），`loginUser.loginType=normal`；表单还有 `captcha` 字段，`GET /login/readCaptcha.action` 返回 `"data:image/png;base64,..."` 的 4 位图形验证码（是否强制尚未验证）。
+- 换票：`POST /teacher/ssoItlmsGxkTicket.action?verificaCode=true&page=` → `ITLMS-<hex>||<sig>`；带 `?ticket=` 访问 `https://jsxx.gds.edu.cn/uc/store/courseRegister` 即拿到学习平台会话，两个域名的 cookie 要一起保存。
+- 课程列表：`GET /uc/store/courseRegister`，页面里是 `showCourse('<课程id>','<课程名>')`。
+- 学习页：`GET /study/course/<课程id>?v=1`（服务端渲染、UTF-8），含 `url1`（CDN 直链 MP4）、`t`（总秒数）、`md5Uid`、`/a_<aid>/...` 上报前缀。
+- 播放注册：`POST /a_<aid>/study/video/user/updateVideoStatus1`（`_method=PUT&id=<课件id>&md5Uid=<uid>&playbackSessionId=<sid>&playbackTakeover=true`）→ `{time, md5}`。
+- 进度心跳：`POST /a_<aid>/study/video/user/<课件id>/updateViewTime1`（`_method=PUT&aid=...&isLimit=false&md5&expires&md5Uid&lastViewTime=<秒>&playbackSessionId`）。
+- 已看秒数：`GET /a_<aid>/study/video/user/getViewTime?id=&md5=&expires=&md5Uid=`，`responseData` 直接就是秒数。
+
+### 踩到的坑
+
+- 三个上报接口必须带 **Referer（指向课程页）**，否则 500 且响应体为空。
+- `isLimit` 只能传 `false` / `true`，传 `"N"` 会 400。
+- 课程列表页、学习页都是 UTF-8（别按 GBK 解，会花屏）。
+
+### 关键结论
+
+进度不靠播放器，`lastViewTime` 直接写服务端。实测把 2640 秒的课报到 2700 秒也被接受，说明可以跳着报、不必真播；默认 `speed=60`、`beat=2`，44 分钟的课约 45 秒报完。课程是否学完看 `getViewTime >= t`。
+
+### 怎么测
+
+1. 「添加账号」平台选「广东公需课」，填手机号 + 密码 → 能拉到课程列表就算登录成功。
+2. 选一门**没学完**的课 → 开始学习 → 看「课程日志」里的进度上报行，再回平台页面核对已学时长。
+3. 登录若报「平台要求图形验证码」，说明密码登录要过验证码，先改用浏览器登录后导入 cookie，再补 OCR。
+
+### 如果没生效
+
+- 400 → 看 Referer 有没有、`isLimit` 是不是 `false`。
+- 500 且空响应 → 少了 Referer，或 `md5Uid` 取错。
+- 课程列表为空 → jsxx 会话失效，重新登录（或等 `renew_session` 换票）。
diff --git a/docs/progress/todo.md b/docs/progress/todo.md
index 2dc7be6..7dd7a69 100644
--- a/docs/progress/todo.md
+++ b/docs/progress/todo.md
@@ -5,3 +5,6 @@
 - 前端补一条自动化回归（Playwright）：空态、账号卡片、测验弹窗、设置保存、平台下拉。
 - 构建检查进 CI：`cd web && pnpm install --frozen-lockfile && pnpm typecheck && pnpm build`，
   并校验 `static/` 与源码同步。
+- 广东公需课：确认 `loginIn.do` 的图形验证码是否强制；若强制，接 OCR 或大模型识图后再自动登录。
+- 广东公需课：用一门**未学完**的课实测进度上报，确认平台端「已学时长 / 已完成」状态跟着变。
+- 广东公需课：视频弹题与课程考核未接入（第一期只做登录 / 课程列表 / 看课）。
diff --git a/metadata.yaml b/metadata.yaml
index c50c86f..943bc0a 100644
--- a/metadata.yaml
+++ b/metadata.yaml
@@ -1,3 +1,3 @@
 name: study-tool
-version: 0.0.1
+version: 0.0.2
 description: 东莞市教师培训平台（jspx.dgjy.net / study.dgjy.net）多账号学习控制台
diff --git a/script/capture-jsglpt.py b/script/capture-jsglpt.py
new file mode 100644
index 0000000..7b991a7
--- /dev/null
+++ b/script/capture-jsglpt.py
@@ -0,0 +1,127 @@
+"""广东省教师继续教育平台抓包工具（一次性取证用，不参与运行）。
+
+用法：
+    python script/capture-jsglpt.py [分钟数]
+
+行为：开一个有头 Chromium（持久 profile，登录态会保留），打开 jsglpt 登录页，
+把你在窗口里操作期间产生的 XHR/fetch 全量落到 data/capture-jsglpt-<时间>.jsonl，
+同时每 3 秒把 cookie 快照写到 data/capture-jsglpt-cookies.json。
+"""
+import json
+import os
+import sys
+import time
+
+from playwright.sync_api import sync_playwright
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+DATA = os.path.join(ROOT, "data")
+LOGIN = "https://jsglpt.gds.edu.cn/login.jsp"
+MINUTES = float(sys.argv[1]) if len(sys.argv) > 1 else 40
+
+os.makedirs(DATA, exist_ok=True)
+STAMP = time.strftime("%Y%m%d-%H%M%S")
+JSONL = os.path.join(DATA, "capture-jsglpt-%s.jsonl" % STAMP)
+COOKIES = os.path.join(DATA, "capture-jsglpt-cookies.json")
+PROFILE = os.path.join(DATA, ".gd-profile")
+
+out = open(JSONL, "a", encoding="utf-8")
+seen = 0
+
+
+def emit(rec):
+    global seen
+    seen += 1
+    rec["n"] = seen
+    out.write(json.dumps(rec, ensure_ascii=False) + "\n")
+    out.flush()
+    print("  %-4s %s %s -> %s" % (rec.get("rt", "doc"), rec.get("method", ""),
+                                  rec.get("url", "")[:150], rec.get("status", "")), flush=True)
+
+
+def on_response(r):
+    try:
+        req = r.request
+        rt = req.resource_type
+        if rt not in ("xhr", "fetch", "document"):
+            return
+        h = {}
+        try:
+            h = dict(r.headers or {})
+        except Exception:
+            pass
+        body = ""
+        # 只留文本类响应，视频/图片跳过
+        if not (r.url or "").lower().split("?")[0].endswith((".mp4", ".m3u8", ".ts", ".png", ".jpg", ".gif", ".woff", ".woff2", ".css", ".js")):
+            try:
+                body = r.text()[:8000]
+            except Exception:
+                body = "<binary>"
+        emit({
+            "t": time.strftime("%H:%M:%S"),
+            "rt": rt,
+            "method": req.method,
+            "url": r.url,
+            "req_headers": dict(req.headers or {}),
+            "req": (req.post_data or "")[:4000],
+            "status": r.status,
+            "resp_headers": h,
+            "resp": body,
+        })
+    except Exception:
+        pass
+
+
+def main():
+    print("抓包输出：%s" % JSONL, flush=True)
+    with sync_playwright() as p:
+        ctx = p.chromium.launch_persistent_context(
+            PROFILE, headless=False,
+            args=["--disable-blink-features=AutomationControlled"],
+            viewport={"width": 1440, "height": 900},
+        )
+        hooked = set()
+
+        def hook(page):
+            if id(page) in hooked:
+                return
+            hooked.add(id(page))
+            page.on("response", on_response)
+            print("[页面] %s" % (page.url or "about:blank"), flush=True)
+
+        ctx.on("page", hook)
+        pg = ctx.pages[0] if ctx.pages else ctx.new_page()
+        hook(pg)
+        try:
+            pg.goto(LOGIN, wait_until="domcontentloaded", timeout=60000)
+        except Exception as e:
+            print("打开登录页失败：%s" % e, flush=True)
+
+        last = 0.0
+        t0 = time.time()
+        while time.time() - t0 < MINUTES * 60:
+            try:
+                now = time.time()
+                if now - last > 3:
+                    last = now
+                    with open(COOKIES, "w", encoding="utf-8") as f:
+                        json.dump(ctx.cookies(), f, ensure_ascii=False, indent=1)
+                ctx.pages[0].wait_for_timeout(1000)
+            except Exception:
+                # 用户把所有页面关掉了，收工
+                break
+        try:
+            with open(COOKIES, "w", encoding="utf-8") as f:
+                json.dump(ctx.cookies(), f, ensure_ascii=False, indent=1)
+        except Exception:
+            pass
+        try:
+            ctx.close()
+        except Exception:
+            pass
+    out.close()
+    print("抓包结束：%d 条 → %s" % (seen, JSONL), flush=True)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/script/test-gdgxk.py b/script/test-gdgxk.py
new file mode 100644
index 0000000..5e3356f
--- /dev/null
+++ b/script/test-gdgxk.py
@@ -0,0 +1,41 @@
+"""广东公需课自测：用浏览器抓到的 cookie 走一遍 列表 → 发现 → 心跳。
+
+用法：PYTHONPATH=src python script/test-gdgxk.py [课件id]
+"""
+import json
+import sys
+import uuid
+
+from study_tool.platforms import gdgxk
+
+cid = sys.argv[1] if len(sys.argv) > 1 else ""
+cookies = json.load(open("data/capture-jsglpt-cookies.json", encoding="utf-8"))
+s = gdgxk.new_session(cookies)
+
+courses = gdgxk.list_courses(s)
+print("课程数：", len(courses))
+for c in courses:
+    print("  ", c["plan_id"], c["name"])
+
+if not cid and courses:
+    cid = courses[0]["plan_id"]
+
+d = gdgxk.discover(s, cid, "")
+print("\n课程：", d["course"])
+print("进度：", d["progress"])
+print("视频：", d["info"])
+cw = d["courseware"][0]
+print("课件：", cw["id"], "| 总长", cw["total"], "| 已看", cw["viewTime"])
+
+sid = uuid.uuid4().hex
+sign = gdgxk._sign(s, {"aid": cw["aid"], "id": cw["id"], "sid": sid}, cw["uid"])
+print("\n播放注册签名：", sign)
+print("服务端已看秒数：", gdgxk._view_time(s, cw, cw["uid"], sign))
+
+r = gdgxk._xhr(s, "POST",
+               gdgxk.X + "/" + cw["aid"] + "/study/video/user/" + cw["id"] + "/updateViewTime1",
+               data={"_method": "PUT", "aid": cw["aid"], "isLimit": "false", "md5": sign["md5"],
+                     "expires": sign["time"], "md5Uid": cw["uid"],
+                     "lastViewTime": str(min(cw["viewTime"] + 60, cw["total"])),
+                     "playbackSessionId": sid})
+print("\n心跳响应：", r.status_code, r.text[:300])
diff --git a/src/study_tool/platforms/__init__.py b/src/study_tool/platforms/__init__.py
index 751ed3e..5454420 100644
--- a/src/study_tool/platforms/__init__.py
+++ b/src/study_tool/platforms/__init__.py
@@ -12,4 +12,5 @@ from study_tool.platforms.base import Platform  # noqa: F401
 
 def load_builtin():
     """导入内置平台，触发注册。幂等，app 启动时调用一次。"""
+    from study_tool.platforms import gdgxk   # noqa: F401
     from study_tool.platforms import guanyixue   # noqa: F401
diff --git a/src/study_tool/platforms/gdgxk.py b/src/study_tool/platforms/gdgxk.py
new file mode 100644
index 0000000..6c7e82f
--- /dev/null
+++ b/src/study_tool/platforms/gdgxk.py
@@ -0,0 +1,271 @@
+"""广东公需课：jsglpt 登录 → jsxx 学习。
+
+登录在 `jsglpt.gds.edu.cn`（账号 / 密码都是 base64），成功后用
+`ssoItlmsGxkTicket.action` 换 ITLMS 票，带票进 `jsxx.gds.edu.cn` 拿学习会话；
+两个域名的 cookie 都留在同一个 requests.Session 里。
+
+学习页是服务端渲染的老系统，视频是 CDN 直链 MP4，进度不靠播放器，靠
+`updateViewTime1` 心跳上报（60 秒一次，`lastViewTime` 是已看秒数），
+`viewTime` 追平视频总长就算学完，所以这里只跑心跳、不真播。
+"""
+import base64
+import json
+import re
+import time
+import uuid
+
+import requests
+
+from study_tool import runtime
+from study_tool.platforms import registry
+from study_tool.platforms.base import Platform
+
+UA = ("Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+      "(KHTML, like Gecko) Chrome/153.0.0.0 Safari/537.36")
+G = "https://jsglpt.gds.edu.cn"
+X = "https://jsxx.gds.edu.cn"
+PLATFORM_ID = "gdgxk"
+
+
+
+def _xhr(s, method, url, **kw):
+    """这些上报接口必须带 Referer（指向课程页），否则直接 500。"""
+    kw.setdefault("headers", {})["X-Requested-With"] = "XMLHttpRequest"
+    return s.request(method, url, timeout=30, **kw)
+
+
+def _expired(msg):
+    """引擎认的是同一个 SessionExpired，这里延迟取避免循环 import。"""
+    from study_tool.engine import SessionExpired
+    return SessionExpired(msg)
+
+
+def _b64(text):
+    return base64.b64encode(str(text).encode("utf-8")).decode()
+
+
+def new_session(cookies=None):
+    s = requests.Session()
+    s.headers.update({"User-Agent": UA, "Accept-Language": "zh-CN,zh;q=0.9"})
+    for c in (cookies or []):
+        if "gds.edu.cn" not in (c.get("domain") or ""):
+            continue
+        s.cookies.set(c["name"], c.get("value") or "",
+                      domain=(c.get("domain") or "").lstrip("."),
+                      path=c.get("path") or "/")
+    return s
+
+
+def dump_cookies(s):
+    return [{"name": c.name, "value": c.value, "domain": c.domain, "path": c.path}
+            for c in s.cookies]
+
+
+def _sso(s):
+    """jsglpt 会话换 ITLMS 票，再把票交给 jsxx 换学习会话。"""
+    r = s.post(G + "/teacher/ssoItlmsGxkTicket.action?verificaCode=true&page=", timeout=30)
+    ticket = (r.text or "").strip()
+    if not ticket.startswith("ITLMS-"):
+        raise _expired("取 ITLMS 票失败：%s" % (ticket[:80] or "空响应"))
+    q = s.get(X + "/uc/store/courseRegister?ticket=" + requests.utils.quote(ticket), timeout=30)
+    if "showCourse" not in q.text and "我的课程" not in q.text:
+        raise _expired("带票进学习平台失败，可能票已失效")
+    return s
+
+
+def login(username, password):
+    s = new_session()
+    s.get(G + "/login.jsp", timeout=30)
+    data = {
+        "loginUser.userName": _b64(username),
+        "loginUser.loginType": "normal",
+        "loginUser.password": _b64(password),
+        "captcha": "",
+    }
+    r = s.post(G + "/loginIn.do", data=data, allow_redirects=False, timeout=30)
+    if r.status_code != 302 or "/default.jsp" not in (r.headers.get("location") or ""):
+        raise RuntimeError("登录失败：账号或密码不对，或平台要求图形验证码（HTTP %s）" % r.status_code)
+    _sso(s)
+    return s, profile(s) or username
+
+
+def renew_session(s):
+    try:
+        _sso(s)
+        return True
+    except Exception:
+        return False
+
+
+def profile(s):
+    html = s.get(G + "/teacher/index.do", timeout=30).text
+    m = re.search(r"你好[，,]?\s*([^\s<]+)\s*老师", re.sub(r"\s+", " ", html))
+    if m:
+        return m.group(1)
+    m = re.search(r'class="user"[^>]*>\s*<[^>]*>\s*([^\s<]{2,20})', html)
+    return m.group(1) if m else ""
+
+
+def list_courses(s):
+    html = s.get(X + "/uc/store/courseRegister", timeout=30).text
+    out, seen = [], set()
+    for cid, name in re.findall(r"showCourse\('([^']+)','([^']*)'\)", html):
+        if cid in seen:
+            continue
+        seen.add(cid)
+        out.append({"plan_id": cid, "project_id": "", "name": name.strip(),
+                    "in_study": True, "platform": PLATFORM_ID})
+    if not out:
+        if "loginBtn" in html or "showLogin" in html:
+            raise _expired("学习平台会话已失效，请重新登录")
+        return []
+    return out
+
+
+def parse_course(html):
+    """从学习页里抠出心跳需要的一切：aid、课件 id、md5Uid、总长、已看秒数。"""
+    d = {}
+    m = re.search(r"/a_([0-9a-f]{32})/study/video/user/", html)
+    if not m:
+        raise RuntimeError("学习页里没有找到课件上报路径，页面结构可能变了")
+    d["aid"] = "a_" + m.group(1)
+    m = re.search(r"updateVideoStatus1'[\s\S]{0,200}?id=([0-9a-f]{32})", html) \
+        or re.search(r"id=([0-9a-f]{32})&md5Uid=", html)
+    if not m:
+        raise RuntimeError("学习页里没有找到课件 id")
+    d["id"] = m.group(1)
+    m = re.search(r"md5Uid=([0-9a-f]{64,128})", html)
+    if not m:
+        raise RuntimeError("学习页里没有找到 md5Uid")
+    d["uid"] = m.group(1)
+    m = re.search(r"\bt\s*=\s*'(\d+)'", html)
+    d["total"] = int(m.group(1)) if m else 0
+    m = re.search(r"url1\s*=\s*'([^']+)'", html)
+    d["video"] = m.group(1) if m else ""
+    d["title"] = (re.search(r"<title>([^<]*)</title>", html).group(1).strip()
+                  if "<title>" in html else "")
+    return d
+
+
+def _sign(s, cw, uid):
+    """updateVideoStatus1：注册播放权，换回心跳签名（time / md5）。"""
+    r = _xhr(s, "POST", X + "/" + cw["aid"] + "/study/video/user/updateVideoStatus1",
+             data={"_method": "PUT", "id": cw["id"], "md5Uid": uid,
+                   "playbackSessionId": cw["sid"], "playbackTakeover": "true"})
+    j = r.json() if r.text.strip().startswith("{") else {}
+    if j.get("responseCode") != "00":
+        raise RuntimeError("注册播放失败：%s" % (r.text[:120] or "空响应"))
+    return j["responseData"]
+
+
+def _view_time(s, cw, uid, sign):
+    r = _xhr(s, "GET", X + "/" + cw["aid"] + "/study/video/user/getViewTime",
+             params={"id": cw["id"], "md5": sign.get("md5"), "expires": sign.get("time"),
+                     "md5Uid": uid})
+    j = r.json() if r.text.strip().startswith("{") else {}
+    rd = j.get("responseData")
+    if isinstance(rd, dict):
+        return int(rd.get("viewTime") or 0)
+    return int(rd or 0)
+
+
+def discover(s, plan_id, project_id, name=""):
+    cid = plan_id
+    page = X + "/study/course/" + cid
+    s.headers["Referer"] = page
+    r = s.get(page + "?v=1", timeout=30)
+    if "loginBtn" in r.text and "md5Uid" not in r.text:
+        raise _expired("学习平台会话已失效，请重新登录")
+    cw = parse_course(r.text)
+    sid = uuid.uuid4().hex
+    cw["sid"] = sid
+    sign = _sign(s, {"aid": cw["aid"], "id": cw["id"], "sid": sid}, cw["uid"])
+    cur = _view_time(s, cw, cw["uid"], sign)
+    done = cw["total"] > 0 and cur >= cw["total"]
+    return {
+        "user": profile(s),
+        "course": name or cw["title"],
+        "progress": "%d/%d 秒" % (cur, cw["total"]),
+        "info": cw["video"],
+        "courseware": [{"id": cw["id"], "title": cw["title"], "href": page,
+                        "total": cw["total"], "viewTime": cur, "done": done,
+                        "aid": cw["aid"], "uid": cw["uid"]}],
+        "feedback": [], "test": [],
+        "platform": PLATFORM_ID, "generation": "gxk",
+        "plan_id": cid, "project_id": "",
+    }
+
+
+def _stopped(key):
+    with runtime.RUNS_LOCK:
+        return (runtime.RUNS.get(key) or {}).get("status") == "stopped"
+
+
+def study(key, s, d, **opts):
+    """只跑心跳上报：每 beat 秒把 lastViewTime 往前推 beat*speed 秒。"""
+    speed = float(opts.get("speed") or 60)
+    beat = int(opts.get("beat") or 2)
+    total = len(d["courseware"])
+    runtime.set_state(key, status="studying", total=total, done=0)
+    for i, cw in enumerate(d["courseware"], 1):
+        if _stopped(key):
+            runtime.log(key, "已按要求停止")
+            return
+        if cw.get("done"):
+            runtime.log(key, "课件 %s 已学完，跳过" % cw["title"])
+            runtime.set_state(key, done=i)
+            continue
+        runtime.log(key, "开始学习：%s（总长 %d 秒，已看 %d 秒）"
+                    % (cw["title"], cw["total"], cw["viewTime"]))
+        _learn(key, s, cw, speed, beat)
+        runtime.set_state(key, done=i)
+    runtime.set_state(key, status="done", done=total)
+    runtime.log(key, "全部课件心跳上报完成")
+
+
+def _learn(key, s, cw, speed, beat):
+    uid, total = cw["uid"], cw["total"]
+    s.headers["Referer"] = cw.get("href") or X
+    sid = uuid.uuid4().hex
+    base = {"_method": "PUT", "aid": cw["aid"]}
+    sign = _sign(s, {"aid": cw["aid"], "id": cw["id"], "sid": sid}, uid)
+    cur = _view_time(s, {"aid": cw["aid"], "id": cw["id"]}, uid, sign)
+    while cur < total:
+        if _stopped(key):
+            runtime.log(key, "已按要求停止")
+            return
+        time.sleep(beat)
+        cur = min(total, cur + int(beat * speed))
+        data = dict(base, isLimit="false", md5=sign.get("md5"), expires=sign.get("time"),
+                    md5Uid=uid, lastViewTime=str(cur), playbackSessionId=sid)
+        r = _xhr(s, "POST", X + "/" + cw["aid"] + "/study/video/user/" + cw["id"] + "/updateViewTime1",
+                 data=data)
+        j = r.json() if r.text.strip().startswith("{") else {}
+        rd = j.get("responseData") or {}
+        if j.get("responseCode") == "00" and rd.get("md5"):
+            sign = {"md5": rd["md5"], "time": rd.get("time")}
+            server = int(rd.get("viewTime") or cur)
+        else:
+            server = _view_time(s, {"aid": cw["aid"], "id": cw["id"]}, uid, sign)
+        runtime.log(key, "进度上报：本地 %d 秒 / 服务端 %d 秒 / 总长 %d 秒" % (cur, server, total))
+        if server < cur:
+            # 服务端没认这个进度，按它的值继续，别把差距越拉越大
+            cur = server
+        if cur >= total:
+            break
+    runtime.log(key, "课件完成：%s（%d 秒）" % (cw["title"], total))
+
+
+PLATFORM = registry.register(Platform(
+    id=PLATFORM_ID,
+    label="广东公需课",
+    study=study,
+    entries=("jsglpt.gds.edu.cn", "jsxx.gds.edu.cn"),
+    session=new_session,
+    dump_cookies=dump_cookies,
+    login=login,
+    list_courses=list_courses,
+    discover=discover,
+    renew_session=renew_session,
+    profile=profile,
+))
diff --git a/todo.md b/todo.md
index 6826bb5..07c1eae 100644
--- a/todo.md
+++ b/todo.md
@@ -2,7 +2,7 @@
 
 ## 待评估：接入广东公需课平台（jsxx.gds.edu.cn）
 
-**状态**：只沉淀思路，不实现。手上没有公需课账号，接口抓不到，代码不动。
+**状态**：已接入，见 `src/study_tool/platforms/gdgxk.py` 与 `OPTIMIZATION.md`；登录 / 课程列表 / 进度心跳已实测，待用未学完课程确认。
 
 ### 背景