Alibabacloud Wxz Website Builder

Dev Tools

Use when building or modifying websites with AI Staff (零号员工/万小智) via Alibaba Cloud OpenAPI. Supports conversation creation, async chat with requirement collection, PRD generation, code generation, and incremental SSE event polling.

Install

openclaw skills install alibabacloud-wxz-website-builder

Category: service

AI Staff Website Builder (零号员工建站)

Validation

mkdir -p output/alibabacloud-wxz-website-builder
for f in skills/ai/service/alibabacloud-wxz-website-builder/scripts/*.py; do
  python3 -m py_compile "$f"
done
echo "py_compile_ok" > output/alibabacloud-wxz-website-builder/validate.txt

Pass criteria: command exits 0 and output/alibabacloud-wxz-website-builder/validate.txt is generated.

Output And Evidence

Save list/summarize outputs under output/alibabacloud-wxz-website-builder/.
Keep conversation IDs and event summaries in each evidence file.

Prerequisites

pip install alibabacloud_tea_openapi>=0.4.4 alibabacloud_tea_util>=0.3.14

Credentials are resolved via the default credential chain (no explicit AK/SK needed).

Authentication

This skill relies on the Alibaba Cloud default credential chain. Do NOT set AK/SK explicitly. The SDK automatically resolves credentials in the following order:

Environment variables: ALIBABACLOUD_ACCESS_KEY_ID / ALIBABACLOUD_ACCESS_KEY_SECRET
Shared config file: ~/.alibabacloud/credentials
RAM role / ECS metadata service (when running on Alibaba Cloud instances)

Region: ALIBABACLOUD_REGION_ID defaults to cn-hangzhou.

How to obtain AccessKey (if user doesn't have one)

If the user has no AccessKey yet, guide them through these steps (see references/ak-setup-guide.md for full details):

Login: Open https://www.aliyun.com and log in (or register)
Create RAM user: Go to https://ram.console.aliyun.com/users → "Create User" → check "OpenAPI Access" → save the AK/SK immediately (Secret is only shown once!)
Grant permissions: Add a custom policy with the following Actions (least-privilege):
- zero2staff:CreateAIStaffConversation
- zero2staff:CreateAIStaffChat
- zero2staff:ListAIStaffChatEvents
- zero2staff:ListAIStaffChatMessages
Configure: Write to ~/.alibabacloud/credentials or set environment variables

CRITICAL: When guiding the user, remind them:

Do NOT use root account AccessKey — always use RAM sub-user
Save the AccessKey Secret immediately — it's only shown once during creation
Never commit AccessKey to git

If the user encounters auth errors, refer to the troubleshooting table in references/ak-setup-guide.md.

Application Lifecycle

The complete flow has 3 phases. Follow them sequentially.

IMPORTANT — Agent-driven polling: The chat command fires the request and returns immediately. The agent then drives the polling loop via the poll command. Between each poll, the agent MUST show the user a progress message so they know what's happening (use progressDetail for rich messages).

Phase 1: Create Conversation
        ↓
Phase 2: Fire requirement chat → poll → HITL → Fire resume → poll → ... → PRD ready
        ↓
Phase 3: Fire code generation → Show link → poll loop with progress → Site ready

Phase 0: Auth Setup

Ensure Alibaba Cloud credentials are configured via the default credential chain (see Authentication section above).

Phase 1: Create Conversation

MUST create a conversation before any chat operation:

CONV=$(python scripts/aistaff_api.py create-conversation --text "做个popmart首页")
CONV_ID=$(echo $CONV | jq -r '.ConversationId')
CHAT_ID=$(echo $CONV | jq -r '.ChatId')
SITE_ID=$(echo $CONV | jq -r '.SiteId')

Returns flat JSON: {ConversationId, SiteId, ChatId, SectionId, BotId, Title}.

Phase 2: Requirement Collection + PRD

This phase collects requirements and generates a PRD. The platform may ask multiple HITL rounds (basic info → features → language, etc.). To keep things fast:

Only the first HITL round should be shown to the user (basic project info).
All subsequent HITL rounds must be auto-filled with the form's default/pre-selected values and resumed immediately — do NOT ask the user.

Step 1: Fire requirement collection

python scripts/aistaff_api.py chat \
  --text "做个popmart首页" \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID

Tell user: "正在分析您的需求，请稍候..."

Step 2: Poll until first HITL form arrives

Call poll every 5 seconds until phase is waiting_for_input:

python scripts/aistaff_api.py poll \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID \
  --last-event-id 0

Between each poll, show the user a progress message based on phase:

processing → "需求分析进行中..."
fetching_reference → "正在获取参考网站信息..."
waiting_for_input → First HITL form arrived, proceed to Step 3.

Step 3: First HITL — collect answers from user

Extract questions from the metaData.arguments of the message.tool event where name: "AskUserQuestion". Present these questions to the user via the AskUserQuestion tool (typically: app name, business description, target users, reference site).

Step 4: Fire resume with `--phase generate_prd`

CRITICAL: On the first HITL resume, always pass --phase generate_prd --user-navigation generate_prd.

python scripts/aistaff_api.py chat \
  --text '{"应用名称": "POP MART 官网", "主营服务": "潮流玩具", "目标用户": "Z世代潮流青年", "参考网站": "无"}' \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID \
  --chat-id $CHAT_ID \
  --chat-status interrupt \
  --phase generate_prd \
  --user-navigation generate_prd \
  --hidden --without-refer

The --text JSON keys MUST match the header values from the form.

Tell user: "已收到您的需求，正在生成产品方案..."

Step 5: Poll loop — auto-fill subsequent HITL rounds until PRD is ready

Poll every 5 seconds. Based on phase / summary, take action:

phase == "waiting_for_input" (another HITL question) → Auto-fill immediately using the answers field from the AskUserQuestion event, then fire resume again. Tell user: "正在完善需求细节..."
phase == "generating_prd" → Tell user: "PRD 方案生成中，请稍候..."
phase == "fetching_reference" → Tell user: "正在获取参考资料..."
summary.chatStatus == "success" + summary.hasPrd == true → PRD ready, proceed to Phase 3.
summary.chatStatus == "fail" → Ask user whether to retry.

# Poll example:
python scripts/aistaff_api.py poll \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID \
  --last-event-id $LAST_EVENT_ID

# Auto-fill example (use the "answers" field from the AskUserQuestion event):
python scripts/aistaff_api.py chat \
  --text '{"核心功能": ["商品展示", "品牌故事", "新闻资讯"]}' \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID \
  --chat-id $CHAT_ID \
  --chat-status interrupt \
  --phase generate_prd \
  --user-navigation generate_prd \
  --hidden --without-refer

Key rule: The platform's AskUserQuestion event always includes an answers field with sensible defaults. For rounds after the first, always use these defaults directly instead of prompting the user.

Phase 3: Code Generation

When PRD is ready:

Step 1: Fire code generation

python scripts/aistaff_api.py chat \
  --text "确认生成应用" \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID \
  --phase generate_code \
  --without-refer

Step 2: Show site link immediately

MUST show before and after code generation:

https://wanwang.aliyun.com/webdesign/home#/ai/manage/prd?conversationId=<CONV_ID>

Tell user: "代码生成已启动，通常需要 2-5 分钟。您可以先通过上面的链接查看项目，我会持续跟踪进度..."

Step 3: Poll loop with progress updates

Poll every 10 seconds. Show the user progress between each poll:

python scripts/aistaff_api.py poll \
  --conversation-id $CONV_ID \
  --biz-id $SITE_ID \
  --last-event-id $LAST_EVENT_ID

Progress messages to show (use progressDetail for rich messages):

phase == "processing" + progressDetail.latestFile exists → "代码生成中...正在写入: {latestFile.semantic}（已生成 {filesWrittenCount} 个文件）"
phase == "processing" + no latestFile → "代码生成中...已处理 N 个事件"
phase == "processing" + progressDetail.lastMessage exists → Also show the assistant's latest message as context
phase == "success" → Site is ready! Show link again.
phase == "fail" → Ask user whether to retry.

Step 4: Handle completion

summary.chatStatus == "success" → Site is ready. Show site link again.
summary.chatStatus == "fail" or summary.hasError == true → Ask user whether to retry. Do NOT auto-retry.

After Initial Generation

Use chat with --conversation-id to modify the site.
Each modification uses the same chat + poll pattern.
Show the site link again after each modification.

Decision Table

`summary.chatStatus`	`summary.hasPrd`	Phase	Action
`interrupt`	—	2 (first round)	Parse AskUserQuestion, collect answers from user, resume with `--phase generate_prd`
`interrupt`	—	2 (subsequent)	Auto-fill with default `answers` from the AskUserQuestion event, resume immediately — do NOT ask user
`success`	`true`	2	PRD ready → proceed to Phase 3
`success`	`false`	3	Site is ready → show link
`fail`	—	any	Ask user whether to retry
`unknown`	—	any	Poll timed out → use `list-messages` fallback

Available Commands

create-conversation

python scripts/aistaff_api.py create-conversation --text "用户问题"

Returns flat JSON: {ConversationId, SiteId, ChatId, SectionId, BotId, Title}.

chat

Fire an async chat message and return immediately. Automatically drains old SSE events before firing. Use poll afterwards to track progress.

python scripts/aistaff_api.py chat --text "description" --conversation-id <ID> --biz-id <ID> [options]

Key parameters:

--text TEXT: Message text or form answers JSON (required)
--conversation-id ID / --biz-id ID: Required identifiers
--chat-id ID + --chat-status interrupt: For HITL resume
--phase {requirement_collect,generate_prd,generate_code}: Phase to enter
--user-navigation TARGET: Navigation target (use generate_prd with --phase generate_prd)
--hidden / --without-refer: Hide message / skip reference context
--verbose: Show debug info

Returns JSON:

{
  "conversationId": "conv-xxx",
  "chatId": "chat-xxx",
  "status": "fired",
  "error": null
}

poll

Single-shot status check — fetches new events once + checks message status, then returns immediately. Designed for agent-driven polling loops where the agent controls the pace and shows progress to the user between calls.

python scripts/aistaff_api.py poll --conversation-id <ID> --biz-id <ID> [--last-event-id N]