HLZD 市场情报报告
把"调研出来的信号"组装成 9 节自包含 HTML 报告,便于发给客户、董事会议或朋友圈。
When to use
调用本 Skill 当用户:
- 跑完
hlzd-b2b-research拿到 JSON,想打包成可视报告 - 跑完
hlzd-buyer-finder拿到买家名单,想附市场分析 - 要给客户/老板看"我做了什么"的市场扫描
- 想在 5 分钟内交付一份视觉 + 数据 都达标的报告
不要调用本 Skill 当:
- 用户只想要一份纯文本摘要 —— 用
hlzd-b2b-research的 Markdown 输出即可 - 用户要做实时仪表盘 —— 这是静态报告(v0.1)
How this skill is invoked
# 全自动:topic + 关键词 → 一键 HTML + Markdown
py scripts/pipeline.py --topic "OCTG casing export to UAE" \
--keywords "OCTG casing" "API 5CT" "Aramco tender" \
--output-html report.html --output-md report.md
# 复用 hlzd-b2b-research 的 JSON
py scripts/pipeline.py --topic "auto" \
--from-research b2b-research-output.json \
--output-html report.html
9-节报告结构
┌── Cover ────────────────────────┐
│ ★ Headline (L1: 56-72px) │
│ Tagline (L2: 16-18px) │
│ 4 KPI cards │
│ 3-step action plan (L1 action) │
│ Meta line (L3: 13px) │
└────────────────────────────────┘
↓
TOC (9 anchored links)
↓
1. Solution Overview
- Recap (16-18px)
- 5-row conclusion table
↓
2. 3 Core Signals
- User pain quote (blockquote)
- Root cause
- Market implication
- Source chips
↓
3. Platform Deep Dive (×4-5 platforms)
- Stat chip
- 5 findings each (with inline source link)
↓
4. Strategic Comparison Table
- 10 dimensions × 3 vendors
↓
5. 3-Step Action Plan
- Green card with options / risk / execution
↓
6. Methodology & Limitations
- 5-step process / 5-dim signal judgement / 7-limitations
- Yellow callout box
↓
7. Sources (grouped by category)
- Sources grid
↓
Footer (HLZD brand + license)
Output schema (input 数据契约)
{
"$schema": "hlzd/market-report/v1",
"meta": { "topic": str, "audience": str, "window_days": int, "language": str },
"cover": {
"headline": str, "tagline": str,
"kpis": [ {"label": str, "value": str, "unit": str?}, ... ],
"three_step_plan": [ {"step": str, "action": str}, ... ]
},
"toc": [ {"id": str, "title": str}, ... ],
"solution_overview": {
"recap": str,
"conclusion_table": [ {"metric": str, "finding": str, "implication": str}, ... ]
},
"three_signals": {
"signals": [
{ "title": str, "user_pain_quote": str, "root_cause": str,
"market_implication": str, "sources": [ {"name": str, "url": str}, ... ] }
]
},
"platform_deep_dive": {
"platforms": [
{ "name": str, "stat": str,
"findings": [ {"headline": str, "body": str, "source_name": str, "source_url": str}, ... ] }
]
},
"strategic_comparison": {
"vendors": [ { "name": str, "attr_1": str, ... }, ... ],
"attributes": [str, ...]
},
"action_plan": {
"actions": [ {"title": str, "options": [str, ...], "risk": str, "execution": str}, ... ]
},
"methodology": {
"five_step": [str, ...], "five_dim": [str, ...], "seven_limitations": [str, ...]
},
"footer_sources": {
"sources": [ {"name": str, "url": str, "category": str}, ... ]
}
}
Voice Contract (LAW 7 条)
继承自 b2b-overseas-market-report spec,落地为可执行检查:
| LAW | 含义 | 实施位置 |
|---|---|---|
| 1 | Body 开头 = "What I learned:" | lib.check_law_1 |
| 2 | 永远用 - 不用 em-dash / en-dash |
lib.normalize_dashes 自动替换 |
| 3 | 引用必须 [name](url) inline link |
lib.md_link |
| 4 | 无 trailing Sources: block(footer 替代) |
lib.check_law_4 |
| 5 | 每个 claim 必须有 URL 或 engagement signal | lib.validate_voice_contract |
| 6 | "best X 2026" 按信号质量排 | heuristic,由 agent 做 |
| 7 | 不假设用户是某个品牌 | lib.validate_voice_contract (brand_assumption) |
输出前 validate_voice_contract() 跑一遍,violations 列表可由 pipeline 选 warn 或 block。
Integration with hlzd-b2b-research
# 跑完 b2b-research → 出 Markdown
py ../hlzd-b2b-research/scripts/run_research.py \
--product "OCTG casing" --markets "UAE Saudi" \
--output-json b2b-output.json
py scripts/pipeline.py --topic "auto" --from-research b2b-output.json \
--output-html report.html --output-md report.md
from_b2b_research() adapter 自动把 b2b-research 的 trade_data / trends / buyers 转成 report schema:
- 每个 market = 一个 signal
- trade_data → platforms
- HS code + 中国份额 → comparison data
Sample outputs
| 规模 | HTML | Markdown |
|---|---|---|
| Standard (auto) | ~14 KB | ~4 KB |
| With data (full) | ~16 KB | ~5 KB |
CSS 内嵌(自包含),无外部字体 / CDN 依赖。Print-friendly:自动翻转为高对比黑白版。
Anti-pattern / Limitations
| 限制 | 处置 |
|---|---|
Brave API 需要 HLZD_BRAVE_API_KEY |
fallback 到 ddgs;都没有则手动输入信号 |
| pip install ddgs 是 v0.1 唯一可选外部依赖 | 需要网络爬取能力 |
| 报告不含 LLM 调用 — signals 是 heuristic 提炼 | 上层 Agent 可接 LLM 增强 from_b2b_research 提炼 3 core signals |
| Voice Contract 检查是 advisory — 不阻断 | 上层 Agent 可硬 fail 在 violation > 0 |
| 多人多语言一次性渲染:v0.1 仅 en 模板 | 二期加 zh / es / ar 多语言模板 |
Versioning
| 版本 | 说明 |
|---|---|
| 0.1.0 | 首版:9 节 HTML + 8 节 Markdown + 从 hlzd-b2b-research 接入 |
升级轨迹见仓库根 VERSIONS.md。
Crafted for B2B industrial exporters — HLZD Cross-Border AI Platform · 2026