Finance News Skill

Attribution: based on finance skill patterns by @@anton-roos.

Fetch recent finance news for Yahoo symbols (stocks, ETFs, indices) using yfinance.

It is optimized for:

• quick headline checks before market decisions
• summarizing top stories by symbol
• avoiding repeated requests with local cache

When to use

Use this skill when the user asks:

• "What are the latest news for AAPL?"
• "Show recent headlines for TSLA."
• "Give me links and publication times for MSFT news."
• "Any major news on ^GSPC today?"

If the user asks for multiple symbols, run the script once per symbol and merge/summarize results.

Common trigger phrases:

• "latest news"
• "stock news"
• "ticker headlines"
• "news for"
• "market news"

Provider strategy

• Default provider: Yahoo Finance via yfinance
• No API key required
• Data quality/availability is best-effort and may vary by symbol
• News freshness and coverage can vary by region, source, and Yahoo response quality

Quick start (how you run it)

These scripts are intended to be run from a terminal. The agent should:

1. ensure dependencies installed
2. run the scripts
3. summarize results cleanly

Install:

cd {baseDir}
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
deactivate

Quick start (agent execution)

Run from this skill directory.

Install dependencies:

pip install -r {baseDir}/requirements.txt

Commands

1) Fetch recent news (JSON)

Examples:

python {baseDir}/scripts/market_news.py AAPL
python {baseDir}/scripts/market_news.py TSLA --limit 15
python {baseDir}/scripts/market_news.py ^GSPC --limit 10

2) Force refresh (skip cache)

python {baseDir}/scripts/market_news.py NVDA --no-cache

3) Tune cache behavior

python {baseDir}/scripts/market_news.py AAPL --ttl 60
python {baseDir}/scripts/market_news.py AAPL --ttl 0 --no-cache

4) Multi-symbol workflow

python {baseDir}/scripts/market_news.py AAPL --limit 8
python {baseDir}/scripts/market_news.py MSFT --limit 8
python {baseDir}/scripts/market_news.py NVDA --limit 8

Then produce one consolidated summary grouped by symbol.

Agent workflow

1. Detect requested symbol(s) and desired depth (limit, default 10).
2. Run the script for each symbol.
3. Parse JSON and extract top items by recency and relevance.
4. For every returned news item, include its URL (url preferred, fallback to link).
5. Return a concise human summary plus per-item links.
6. If there are zero items, say so clearly and suggest retry (--no-cache) or another symbol.

Output expectations

• Script output is JSON with: symbol, source, fetched_at_*, count, items, note.
• Each item usually includes: title, publisher, published_at_unix, published_at_utc, url, link, summary.
• User-facing response should include:
  • top headlines (up to requested limit)
  • publish time (UTC)
  • source/publisher
  • direct URL/link for each headline

Mandatory rule:

• Do not output a headline without a URL.
• For each item, use url first; if missing, use link.
• If both are missing, write URL not available for that item.

Recommended item format:

- <title>
	Publisher: <publisher>
	Published (UTC): <published_at_utc>
	URL: <url or link>

Suggested response structure:

1. One-line market/news context (if obvious).
2. Bulleted top headlines.
3. Caveat if coverage is sparse or stale.

Safety / correctness

• Do not claim coverage is complete or real-time.
• Mention Yahoo/yfinance is unofficial and best-effort.
• If Yahoo is unreachable or rate-limited, report failure and suggest retrying with lower frequency.
• Avoid trading advice phrasing; present informational summaries.

Failure handling

• count = 0: return "No recent items found" and suggest --no-cache retry.
• Network/provider error: surface error briefly, then propose retry.
• Repeated rate-limit symptoms: increase cache TTL and reduce request frequency.