dk-wikibot

Files (2)

• SKILL.md
• api.md · references/

SKILL.md (first lines)

# Pywikibot Skill

Pywikibot wraps the MediaWiki API. For **read-only** tasks (the common case), no credentials are
needed — set `PYWIKIBOT_NO_USER_CONFIG=1`.

Always write a standalone Python script and execute it via `bash_tool`. Do not import pywikibot
interactively — it makes blocking network calls and must run as a script.

## Setup

```bash
uv add pywikibot mwparserfromhell
export PYWIKIBOT_NO_USER_CONFIG=1
```

Every script must start with:

```python
import os; os.environ['PYWIKIBOT_NO_USER_CONFIG'] = '1'
import pywikibot
```

---

## Step-by-step instructions

### 1. Fetch a Wikipedia article

```python
import os; os.environ['PYWIKIBOT_NO_USER_CONFIG'] = '1'
import pywikibot

site = pywikibot.Site('en', 'wikipedia')
page = pywikibot.Page(site, 'Grammy Award for Best Rock Song')
print(page.text[:3000])  # raw wikitext — inspect this first before parsing
```

Always inspect `page.text` before writing any parser. Wikitext structure varies per article.

### 2. Parse wikitext

```python
import mwparserfromhell
wikicode = mwparserfromhell.parse(page.text)
templates = wikicode.filter_templates()  # e.g. {{won}}, {{nom}}, {{infobox}}
```

Alternatively, use regex directly on `page.text`:

```python
import re
year_sections = re.split(r'\n==\s*(\d{4})\s*==\n', page.text)
links = re.findall(r'\[\[([^\|\]]+)(?:\|[^\]]+)?\]\]', block)
```

### 3. Iterate a Wikipedia category

```python
import os; os.environ['PYWIKIBOT_NO_USER_CONFIG'] = '1'
import pywikibot

site = pywikibot.Site('en', 'wikipedia')
cat = pywikibot.Category(site, 'Category:Grammy Award for Best Rock Song')
for page in cat.articles(total=50):
    print(page.title())
```

Use `cat.subcategories()` for subcats, `cat.members()` for everything.

### 4. Query Wikidata via SPARQL

Best for clean structured data (winners, dates, relationships). Find the QID for any entity at
https://www.wikidata.org/wiki/Special:Search first, then use it in the query.

```python
import os; os.environ['PYWIKIBOT_NO_USER_CONFIG'] = '1'
from pywikibot.data import sparql

QUERY = """
SELECT ?year ?songLabel ?artistLabel WHERE {
  ?award wdt:P31 wd:Q1377714 .       # instance of Grammy Award for Best Rock Song
  ?award wdt:P585 ?date .
  BIND(YEAR(?date) AS ?year)
  FILTER(?year >= 2010)
  SERVICE wikibase:label { bd:serviceParam wikibase:language "en". }
}
ORDER BY ?year
"""
results = sparql.SparqlQuery().select(QUERY)
for row in results:
    print(row)
```

### 5. Fetch a Wikidata item directly

```python
import os; os.environ['PYWIKIBOT_NO_USER_CONFIG'] = '1'
import pywikibot

site = pywikibot.Site('en', 'wikipedia')
page = pywikibot.Page(site, 'Douglas Adams')
item = pywikibot.ItemPage.fromPage(page)
item.get()
print(item.labels.get('en'))
print(item.claims)  # dict of property_id -> list of Claims
```

---

## Decision guide

| Task | Approach |
|------|----------|
| Award winner lists, formatted tables | Wikipedia `page.text` + wikitext parsing |
| Clean structured facts (dates, IDs, relationships) | Wikidata SPARQL or `ItemPage` |
| Category membership | `Category.articles()` |
| Multiple entities at once | Wikidata SPARQL (most efficient) |

---

From dushyantkhosla/agent-skills, Apache-2.0.

More from dushyantkhosla/agent-skills