Law Download

Downloads official Dutch legal publications from government sources (national and local) and converts them to the regelrecht YAML format.

Pipeline: /law-download (this skill) → /law-interpret (adds machine_readable logic)

Use this skill first to create a text-only YAML file, then use /law-interpret to add machine-readable execution logic to it.

What This Skill Does

Searches for Dutch regulations in government databases:
- BWB (Basiswettenbestand) - National laws and regulations
- CVDR (Centrale Voorziening Decentrale Regelgeving) - Local regulations
Presents search results with identifiers (BWBR, CVDR) and metadata
Downloads the XML from the official repository
Parses the XML to extract articles and metadata
Converts to YAML format with text only (no machine_readable sections)
Saves to corpus/regulation/nl/{layer}/{law_id}/{date}.yaml

Supported Regulation Types

National Level (BWB)

Wetten (formal laws)
AMvB (Algemene maatregel van bestuur)
Ministeriele regelingen
Koninklijk besluiten
Beleidsregels (implementation policies)

Local Level (CVDR)

Gemeentelijke verordeningen (municipal ordinances)
Provinciale verordeningen (provincial ordinances)
Waterschapsverordeningen (water board ordinances)
Uitvoeringsbeleid (implementation policies)

Important Constraints

Keep legal text EXACTLY as it appears in the source (preserve all formatting, markdown links)
Do NOT convert monetary amounts (keep as written: "€795,47" not "79547")
Do NOT add any machine_readable sections (leave completely empty)
Extract ALL articles - do not skip any
Preserve article structure and numbering exactly

Step-by-Step Instructions

Step 1: Determine Database to Search

Based on the regulation type, choose the appropriate database:

For National Regulations: Use x-connection=BWB

Wetten, AMvB, ministeriele regelingen, koninklijk besluiten, beleidsregels

For Local Regulations: Use x-connection=CVDR

Gemeentelijke verordeningen, provinciale verordeningen, uitvoeringsbeleid

If unclear: Search both databases and combine results

Step 2: Search for the Regulation

API Endpoint (BWB):

https://zoekservice.overheid.nl/sru/Search?operation=searchRetrieve&version=1.2&x-connection=BWB&query={QUERY}&maximumRecords=10

API Endpoint (CVDR):

https://zoekservice.overheid.nl/sru/Search?operation=searchRetrieve&version=1.2&x-connection=CVDR&query={QUERY}&maximumRecords=10

Query Construction:

For title: dcterms.title%20any%20"{name}"
For BWBR ID: dcterms.identifier=={BWBR_ID}
For CVDR ID: dcterms.identifier=={CVDR_ID}
For municipality: overheidcvdr.organisatietype==gemeenten AND dcterms.title%20any%20"{name}"
URL encode spaces as %20

Examples:

# Search BWB
https://zoekservice.overheid.nl/sru/Search?operation=searchRetrieve&version=1.2&x-connection=BWB&query=dcterms.title%20any%20"zorgtoeslag"&maximumRecords=10

# Search CVDR for municipal regulations
https://zoekservice.overheid.nl/sru/Search?operation=searchRetrieve&version=1.2&x-connection=CVDR&query=dcterms.title%20any%20"afvalstoffenverordening"&maximumRecords=10

# Search specific municipality
https://zoekservice.overheid.nl/sru/Search?operation=searchRetrieve&version=1.2&x-connection=CVDR&query=overheidcvdr.organisatietype==gemeenten%20AND%20dcterms.creator==Amsterdam%20AND%20dcterms.title%20any%20"afval"&maximumRecords=10

Step 3: Parse Search Results

Extract from the XML response:

For BWB results:

<dcterms:title> - Law title
<dcterms:identifier> - BWBR ID (e.g., "BWBR0018451")
<dcterms:type> - Type (wet, AMvB, ministeriele regeling, etc.)
<overheidbwb:geldigheidsdatum> - Effective date

For CVDR results:

<dcterms:title> - Regulation title
<dcterms:identifier> - CVDR ID (e.g., "CVDR123456_1")
<dcterms:creator> - Municipality/organization name
<overheidcvdr:organisatietype> - Type (gemeenten, provincies, waterschappen)
<dcterms:issued> - Issue date

Present results to user with format:

Found {N} results:

1. {Title}
   ID: {BWBR_ID or CVDR_ID}
   Type: {Type}
   Organization: {Creator} (if CVDR)
   Latest version: {Date}

2. ...

Ask user: "Which regulation would you like to download? (Enter number or ID)"

Step 4: Download XML Files

Once user selects a regulation, download the appropriate XML files:

For BWB (National) Regulations:

A. WTI File (Metadata):

https://repository.officiele-overheidspublicaties.nl/bwb/{BWBR_ID}/{BWBR_ID}.WTI

B. Toestand File (Legal Text):

https://repository.officiele-overheidspublicaties.nl/bwb/{BWBR_ID}/{DATE}/xml/{BWBR_ID}_{DATE}.xml

For CVDR (Local) Regulations:

Download the CVDR XML directly:

https://repository.overheid.nl/{CVDR_PATH}/xml/{CVDR_ID}.xml

Where CVDR_PATH is extracted from the search results <gzd:resourceIdentifier>.

If date not specified by user, use the latest version from search results.

Use WebFetch or Bash with curl to download these files.

Step 5: Parse Metadata XML

Extract the following from the WTI XML:

XML Namespaces:

xmlns:bwb-dl="http://www.geonovum.nl/bwb-dl/1.0"

Fields to Extract:

<bwb-dl:bwb-id> → bwb_id (top-level field)
<bwb-dl:soort> → Map to regulatory_layer:
- "wet" → "WET"
- "AMvB" → "AMVB"
- "ministeriele regeling" → "MINISTERIELE_REGELING"
- "beleidsregel" → "BELEIDSREGEL"
- "koninklijk besluit" → "KONINKLIJK_BESLUIT"
- See schema for full enum list
<bwb-dl:citeertitel> or <bwb-dl:officiele-titel> → name (and slugified for directory name)
First <bwb-dl:intrekking datum="..."> → valid_from
<bwb-dl:publicatiedatum> → publication_date

Step 6: Parse Legal Text XML for Articles

XML Namespaces:

xmlns:bwb="http://www.overheid.nl/2011/BWB"

Article Structure in XML:

<artikel eId="chp_X__art_Y" wId="BWBR..." status="goed">
  <kop>
    <label>Artikel</label>
    <nr status="officieel">Y</nr>
  </kop>
  <lid eId="..." status="goed">
    <lidnr status="officieel">1</lidnr>
    <al>Legal text here...</al>
  </lid>
  <lid>
    <lidnr>2</lidnr>
    <al>More text...</al>
  </lid>
</artikel>

Extraction Logic:

Find all <artikel> elements
For each article:
- Extract <nr> as article number
- Collect ALL <lid> (paragraphs) and <al> (text blocks)
- Convert to markdown format:
  - Keep paragraph structure
  - Convert <nadruk> to bold or italic
  - Convert <extref> to markdown links [text](url)
- Preserve exact formatting and line breaks

Generate article URL:

https://wetten.overheid.nl/{BWBR_ID}/{DATE}#Artikel{NUMBER}

Step 7: Generate YAML File

Target Structure:

$schema: https://raw.githubusercontent.com/NederlandseDigitaleDienst/regelrecht/refs/tags/schema-v0.8.8/schema/v0.8.8/schema.json
$id: "{LAW_SLUG}"
name: "{LAW_TITLE}"
regulatory_layer: "{MAPPED_LAYER}"
publication_date: "{YYYY-MM-DD}"
valid_from: "{YYYY-MM-DD}"
url: "https://wetten.overheid.nl/{BWBR_ID}/{DATE}"
bwb_id: "{BWBR_ID}"

articles:
  - number: "{ARTICLE_NUMBER}"
    text: |
      {MARKDOWN_TEXT}
    url: "https://wetten.overheid.nl/{BWBR_ID}/{DATE}#Artikel{NUMBER}"
  - number: "{NEXT_ARTICLE}"
    text: |
      {MORE_TEXT}
    url: "..."

Important:

Do NOT include machine_readable sections
Keep text as-is (no eurocent conversion)
Include ALL articles from the law
Use proper YAML multiline string format (|) for text
All identifiers (bwb_id, url) are top-level, not nested

Step 8: Save File

Directory Structure:

corpus/regulation/nl/{regulatory_layer_lowercase}/{law_id}/{valid_from}.yaml

Example:

corpus/regulation/nl/wet/wet_op_de_zorgtoeslag/2025-01-01.yaml
corpus/regulation/nl/ministeriele_regeling/regeling_standaardpremie/2025-01-01.yaml

Create directories if they don't exist.

Step 9: Validate YAML Against Schema (with repair loop)

The generated YAML has to pass just validate. The schema is the single source of truth.

just validate {FILE_PATH}

If validation fails — repair (up to 3 rounds):

Read the error output carefully, identify broken fields/structure
Fix the YAML with the Edit tool
Re-run just validate
If still failing after 3 rounds, stop and report the errors to the user

Common validation issues:

Missing required fields (regulatory_layer, publication_date, url, articles)
Missing bwb_id for national laws (WET, AMVB, MINISTERIELE_REGELING, GRONDWET)
Incorrect regulatory_layer enum value
Malformed YAML syntax (bad indentation, unescaped special characters in text)
Invalid date formats (must be YYYY-MM-DD)
Articles missing number, text, or url fields

Do NOT proceed to Step 10 with invalid YAML. A file that fails validation cannot be used by /law-interpret in the next step.

Step 10: Confirm with User

Report:

✓ Downloaded and converted {REGULATION_TITLE}
  ID: {BWBR_ID or CVDR_ID}
  Type: {Type}
  Articles: {COUNT}
  Saved to: {FILE_PATH}
  ✅ Schema validation: PASSED

The YAML file contains the legal text only.
To add machine-readable execution logic, run /law-interpret on this file.

Error Handling

If search returns no results:

Suggest alternative search terms
Ask user if they have the BWBR ID directly

If XML download fails:

Check if date exists (try other dates from manifest)
Verify BWBR ID is correct
Provide direct URL for user to check in browser

If XML parsing fails:

Report which XML element caused the issue
Save raw XML to temp file for manual inspection
Ask user if they want to continue with partial data

Tips for Success

Always download BOTH WTI and Toestand files
Handle XML namespaces correctly
Preserve exact text formatting (spaces, line breaks)
Generate human-readable directory name slugs (lowercase, underscores — e.g., wet_op_de_zorgtoeslag)
Double-check all articles are included (count them)
Validate YAML syntax before saving

law-download

Files4

SKILL.md.claude/skills/law-download/SKILL.md