Agent Skills: Aggregating Crypto News

Aggregating Crypto News

Overview

Fetch configured RSS feeds on demand, filter dated articles by coin/category and UTC time window, and rank them using keyword and recency scores. Source availability and coverage vary; the tool reports failed and disabled feeds rather than implying every source was checked successfully.

Prerequisites

The user prepares the Python environment before invoking this skill.

1. Python 3.8+ installed
2. Dependencies: pip install feedparser requests pyyaml
3. Internet connectivity for RSS feed access

Instructions

1. Assess user intent - determine filters needed:

   • General news: no filters, use defaults
   • Coin-specific: extract symbol (BTC, ETH, etc.)
   • Category-specific: extract category (defi, nft, regulatory, etc.)
   • Time-specific: extract window (1h, 4h, 24h, 7d)

2. Run the aggregator with appropriate filters:

   # Default scan (top 20, past 24h, relevance sorted)
   python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py
   
   # Coin-specific scan
   python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --coin BTC --period 4h
   
   # Category filter
   python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --category defi --top 30
   
   # Multiple filters
   python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --coin ETH --category defi --period 24h --top 15
   

3. Export results for downstream processing:

   python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --format json --output news.json
   

4. Present results to the user:

   • Show source, title, age, and relevance score
   • Highlight market-moving keywords if present
   • Provide links for full articles
   • Summarize configured/disabled, fetched/failed sources and articles found
   • Report feed failures, including when other feeds return matching articles
   • A zero-match query is distinct from an all-feed failure (exit 1)
   • JSON meta.feed_health records per-source parsed/window/undated counts and latest publication; CSV keeps diagnostics on stderr
   • Undated and future-dated items are excluded from the requested time window

Output

Table showing articles ranked by relevance score (0-100) based on market-moving keyword detection, source authority, and recency:

CRYPTO NEWS AGGREGATOR — [generated UTC time]
[rank] [source] [headline] [age] [score]
Summary: [shown] articles | [attempted] sources | [matched] matches
Feed health: [fetched] fetched | [failed] failed | [disabled] disabled
[source]: failed — [HTTP/timeout/parse reason, when applicable]

This is a layout example, not a record of actual headlines or coverage. RSS publication dates govern the time window, not fetch time or a claimed live scan. The registry retains disabled URLs and dated reasons for review; re-enable a source only after checking its current feed and any publisher usage terms.

Error Handling

Use Read to inspect references/errors.md for comprehensive error handling.

Examples

Filtering patterns for common news monitoring scenarios:

# Latest crypto news (defaults)
python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py

# Bitcoin news from past 4 hours
python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --coin BTC --period 4h

# DeFi category news
python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --category defi

# High-relevance news only
python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --min-score 70 --top 10

# Multiple coins
python ${CLAUDE_SKILL_DIR}/scripts/news_aggregator.py --coins BTC,ETH,SOL

Resources

• references/implementation.md - CLI options, categories, JSON format, advanced filtering
• references/errors.md - Comprehensive error handling
• references/examples.md - Detailed usage examples
• ${CLAUDE_SKILL_DIR}/config/sources.yaml - Full source registry
• CoinDesk RSS: https://www.coindesk.com/arc/outboundfeeds/rss/
• feedparser docs: https://feedparser.readthedocs.io/