news-sentiment

name	news-sentiment
description	Analyze the sentiment of financial news and aggregate coverage into key themes. Use when you need to assess market sentiment, gauge media coverage, or decide whether news about a stock is bullish, bearish, or neutral.

News Sentiment Skill

Purpose

This skill scores the sentiment of financial news text and groups coverage into common themes (earnings, products, market performance, management, regulation). It helps turn a batch of raw headlines or articles into a quick read on market mood.

When to Use

Use this skill when the user (or a coordinating agent):

• Wants to know the overall sentiment of recent news ("bullish or bearish?")
• Needs to summarize a set of headlines or articles
• Asks about media coverage, themes, or market perception of a stock

Capabilities

• Classify sentiment as Positive / Negative / Neutral with a numeric score
• Count positive vs. negative signals
• Cluster news items into key themes
• Optionally export results to JSON

Workflow

1. Gather News Text

Collect the news you want to analyze. If you used the Tavily Search MCP, save the returned headlines/snippets to a file (one item per line) in tmp/.

2. Run the Helper Script

The helper script analyze_sentiment.py does the scoring and theme aggregation:

uv run python .claude/skills/news-sentiment/analyze_sentiment.py \
  --input tmp/news.txt \
  --output tmp/sentiment.json

You can also pass text directly instead of a file:

uv run python .claude/skills/news-sentiment/analyze_sentiment.py \
  --text "Company beats earnings; analysts upgrade to buy"

Arguments:

• --input (optional): Path to a text file of news, one item per line
• --text (optional): News text passed directly (use instead of --input)
• --output (optional): Path to write JSON results (e.g., tmp/sentiment.json)

Provide either --input or --text.

3. Interpret the Results

• Positive (score > 2): coverage skews bullish
• Negative (score < -2): coverage skews bearish
• Neutral (otherwise): mixed or balanced coverage

Use the theme counts to highlight what the news is actually about (e.g., mostly earnings vs. mostly regulatory concerns).

Output Format

When --output is supplied, the script writes JSON like:

{
  "sentiment": {
    "classification": "Positive",
    "score": 4,
    "positive_signals": 5,
    "negative_signals": 1
  },
  "total_items": 8,
  "themes": {
    "Earnings & Financial Results": 4,
    "Product & Innovation": 2
  }
}

Important Notes

• Sentiment is a fast, rule-based keyword heuristic, not a market signal on its own
• Combine it with the actual headlines for context before drawing conclusions
• Use absolute paths under tmp/ for input/output files
• Use uv run python to execute (not plain python)