Daily News Portal (Prasowka)

Generate a daily news portal as a single HTML file.

Parameters

• OUTPUT: \x3Cworkspace>/canvas/prasowka-{YYYYMMDD}.html
• DATA_DIR: \x3Cworkspace>/skills/prasowka/data
• REFS_DIR: \x3Cworkspace>/skills/prasowka/references

Steps

Step 1: Initialize

mkdir -p \x3Cworkspace>/canvas \x3Cworkspace>/skills/prasowka/data
DATE=$(date +%Y%m%d)
SEEN_URLS_FILE="$DATA_DIR/seen_urls.json"

Step 2: Load Configuration

Read $REFS_DIR/topics.md — format:

ai-models: 15
ai-tools: 10
ai-video: 15
...

Step 3: Fetch News per Topic

Launch a subagent for each topic:

def fetch_topic_news(topic, limit):
    result = subprocess.run([
        "python3", "scripts/fetch_news.py",
        "--topic", topic,
        "--limit", str(limit)
    ], capture_output=True, text=True)
    
    if result.returncode != 0:
        return web_search(f"latest {topic} news {limit}")
    
    return json.loads(result.stdout)

Step 4: Filter URLs

def filter_new_urls(articles, seen_urls):
    return [a for a in articles if a['url'] not in seen_urls]

Step 5: Fetch & Summarize per Article

Launch a subagent for each article:

def summarize_article(url):
    content = web_fetch(url, extract_text=True)
    if not content or len(content) \x3C 100:
        return None
    summary = llm_summarize(content[:3000])
    return summary

Step 6: Generate HTML

Use template with dark/light toggle, 18 categories, 2-3 sentence summaries.

Step 7: Save & Update

• Save HTML to canvas
• Update seen_urls.json
• Run prasowka-guardian validation

Error Handling

• fetch_news.py fails → fallback to web_search
• Article unavailable → skip
• Summary empty → use first 3 sentences
• Don't stop — keep going

Requirements

• scripts/fetch_news.py
• references/topics.md
• references/format.md

If this saved you time: ☕ PayPal.me/nerudek