cn-law-hub

Legal Databases Overview

This skill supports three legal databases:

National Laws and Regulations Database (国家法律法规数据库)

Official database: https://flk.npc.gov.cn. Maintained by NPC Standing Committee.

NPC helper scripts: use scripts/download.py for law-level search/download and scripts/article_search.py for article-level keyword extraction across laws.

Environment Selection

This skill supports multiple agent environments. Read the adapter for your environment first:

Kimi Agent and Claude Code via kimi-webbridge intentionally share the same adapter because their browser-operation semantics are the same.

Project Setup

Install Python dependencies once. Old .doc parsing is mainly needed for some NPC legacy regulation files; DOCX parsing uses the Python stdlib.

pip install -r requirements.txt
# Optional: for old .doc format support (some regulations use pre-2007 Word format)
apt-get install antiword catdoc  # Linux
brew install antiword catdoc     # macOS

Page layout and browser automation details are in references/page_structure.md.

Quick Reference

API-First Search → Download Workflow (Recommended)

The agent must decide the search strategy before calling the API. Do not default to fuzzy for every query:

• •Known title / specific regulation (e.g. "物业管理条例", "北京市生活垃圾管理条例"): use --exact (title + exact match) to reduce noise
• •Broad topic / unsure of title (e.g. "出租车", "环境保护"): omit --exact to use fuzzy search
• •Ambiguous query: ask the user whether they want exact title match or broad fuzzy match, or run both and compare

# 1. Fuzzy search by topic
python scripts/download.py --search "出租车" --size 100

# 2. Exact title search (less noise)
python scripts/download.py --search "物业管理条例" --exact --size 100

# 3. Search full text of laws
python scripts/download.py --search "违约金" --range content --size 50

# 4. Filter by effective status
python scripts/download.py --search "出租车" --status 3 --size 50

# 5. Get signed download URLs only (for local batch download)
python scripts/download.py --search "出租车" --urls-only --size 100 > urls.json

# 6. Get metadata
python scripts/download.py --info {bbbs_id}

# 7. Download file
python scripts/download.py --download {bbbs_id} --format docx output.doc

Query Individual Articles (Preview / Article Lookup)

For retrieving specific articles instead of full files:

# Preview law structure — shows article count + numbering pattern + first 20 articles
python scripts/download.py --preview {bbbs_id}

# Query by article number (supports multiple formats, auto-converts)
python scripts/download.py --article {bbbs_id} "第三十八条"   # Chinese
python scripts/download.py --article {bbbs_id} "第38条"      # Arabic
python scripts/download.py --article {bbbs_id} "38"          # Number only

# Grep keyword across all articles of one law
python scripts/download.py --article {bbbs_id} --grep "经济补偿"

If --article misses, read the detected numbering hint and run --preview before retrying.

Search API

POST https://flk.npc.gov.cn/law-search/search/list

Working payload:

{
  "searchRange": 1, "searchType": 2, "searchContent": "出租车",
  "pageNum": 1, "pageSize": 100,
  "orderByParam": {"order": "-1", "sort": ""},
  "flfgCodeId": [], "zdjgCodeId": [], "sxx": [], "gbrq": [], "sxrq": [], "gbrqYear": [],
  "xgzlSearch": false
}

Use the same strategy as the CLI workflow: exact or near-exact titles use searchRange=1 + searchType=1; broad topic searches use searchType=2. Full parameters are in references/api_reference.md.

Browser fallback: Use browser automation only when the API/script fails, when UI-only advanced search is required, or when the user explicitly requests UI operation; read references/page_structure.md first.

Detail API (for metadata)

GET https://flk.npc.gov.cn/law-search/search/flfgDetails?bbbs={bbbs_id}

Returns: title, dates, category, status, and ossFile paths.

Use bundled script: python scripts/download.py --info {bbbs_id}

Download API

GET https://flk.npc.gov.cn/law-search/download/pc?format={docx|pdf}&bbbs={bbbs_id}

Returns a signed OSS URL in data.url. Use that URL to download the actual file.

python scripts/download.py --download {bbbs_id} --format docx output.doc

Note: The ossFile paths from the detail API are not directly downloadable. The site serves the SPA index page for those URLs. Always use the download API or browser download button.

Article-Level Search (scripts/article_search.py)

For finding specific articles that contain a keyword (not just which laws):

# Search across laws whose titles contain keyword, find matching articles
python scripts/article_search.py "违约金" --max-laws 5

# Search across laws whose full text contains keyword
python scripts/article_search.py "违约金" --range content --max-laws 5

# Show surrounding context (1 article before/after each match)
python scripts/article_search.py "抵押权" --range content --max-laws 3 --context 1

# Search within a specific law only
python scripts/article_search.py "善意取得" --law "民法典" --context 0

# JSON output for further processing
python scripts/article_search.py "违约金" --max-laws 3 --json

Progressive Batch Retrieval: use --offset N to skip already-processed laws, or --resume to auto-skip cached ones:

# Batch 1: process first 5 laws
python scripts/article_search.py "违约金" --range content --max-laws 5
# → "已处理: 5/342 部法规"

# Batch 2: skip first 5 laws, then process laws 6-10
python scripts/article_search.py "违约金" --range content --max-laws 5 --offset 5

# Or: use --resume to skip already-processed laws and continue with new ones
python scripts/article_search.py "违约金" --range content --max-laws 5 --resume

Authority / Region Categorization

Use region_classifier.py for province/city/level classification when issuing authority names (zdjgName) are irregular; prefer official zdjgfl codes from GET /law-search/search/enumData when available.

python scripts/download.py --search "物业管理条例" --urls-only --size 100 > urls.json
python scripts/region_classifier.py --classify < urls.json > classified.json
python scripts/region_classifier.py --matrix matrix.csv < classified.json

NPC Rate Limiting

Auto mode (default) picks the mode by estimated request count:

Override: --rate-limit fixed/adaptive/off/N (N = custom req/s).

NPC Cache Management

Local file cache enabled by default (~/.cache/npc-law-db/). Use --cache-stats / --cache-clear / --no-cache.

NPC Status Code Mapping (sxx field)

Filter by status: --status 3 (only current), --status 3,4 (current + upcoming)

Key NPC concepts: bbbs is the unique law ID used in detail URLs and script commands; sxx is the effective-status field (1=已废止, 2=已修改, 3=现行有效, 4=尚未生效).

When the task requires sorting, category codes, download options, advanced search, batch UI actions, or pagination parameters, read references/api_reference.md.

State Council Rules Database (国家规章库)

Official database: https://www.gov.cn/zhengce/xxgk/gjgzk/. Maintained by the State Council.

Quick Start

# Search department rules
python scripts/gov_rules_crawler.py --search "管理办法" --categories 部门规章 --size 20

# Search local government rules
python scripts/gov_rules_crawler.py --search "物业管理" --categories 地方政府规章 --size 20

# Get metadata for a specific page
python scripts/gov_rules_crawler.py --info "https://www.gov.cn/zhengce/202606/content_7073180.htm"

# Download full pages and attachments
python scripts/gov_rules_crawler.py --categories 部门规章 --size 10 --download

Output

Output includes summary.json, metadata.jsonl/csv, stats_report.json/md, logs/, and downloaded files under files/{rule_title}/.

Notes

• •The Athena API requires dynamic RSA authentication discovered from the frontend JS bundle.
• •Auth parameters may expire during long runs; restart if you receive 401/403.
• •Read references/gov_rules_api_reference.md for the full auth and API field mapping.

Ministry of Foreign Affairs Treaty Database (外交条约库)

Official database: https://treaty.mfa.gov.cn/web/. Maintained by the Ministry of Foreign Affairs.

Quick Start

# Search bilateral treaties
python scripts/treaty_crawler.py --collections 双边 --search "上海合作组织" --size 20

# Search multilateral treaties
python scripts/treaty_crawler.py --collections 多边 --search "人权" --size 20

# Get metadata for a specific treaty page
python scripts/treaty_crawler.py --info "https://treaty.mfa.gov.cn/web/detail1.jsp?objid=1531876373617"

# Download treaty preview PDFs
python scripts/treaty_crawler.py --collections 双边 --size 5 --download

Output

Output includes summary.json, metadata.jsonl/csv, stats_report.json/md, logs/, and preview PDFs under files/{treaty_title}_{collection}/.

Notes

• •This is a pure HTML site; parsing relies on BeautifulSoup.
• •Collections: 全部 (all), 双边 (bilateral), 多边 (multilateral).
• •Read references/treaty_api_reference.md for HTML structure and field extraction patterns.

Script Reference Summary

Attribution

Special thanks to [Li2zon3]for the [law-crawler-unified]project.