Last30Days Research & Prompting
Introduction
In the fast-moving world of AI, yesterday’s prompting techniques are often today’s outdated methods. Last30Days is a powerful Claude Code skill designed to bridge the gap between static training data and the “now.” It researches any topic across Reddit, X (Twitter), and the Web specifically from the last 30 days to surface what people are actually discussing, recommending, and debating right now.
Whether you are looking for the latest photorealistic prompting techniques, hunting for the best new Claude Code skills, or staying updated on breaking AI news, Last30Days transforms raw social sentiment into actionable, copy-paste-ready prompts.
Key Features
- Multi-Source Intelligence: Aggregates data from Reddit threads, X posts, and the broader web.
- Engagement-Weighted Synthesis: Uses a “Judge Agent” to prioritize insights with high engagement (upvotes, likes, and shares).
- Intent-Aware Research: Automatically categorizes your request into Prompting, Recommendations, News, or General research.
- Targeted Output: Generates prompts specifically formatted for your tool of choice (e.g., Midjourney, ChatGPT, Claude, or Nano Banana Pro).
- Flexible API Modes: Works in Full Mode (with keys), Partial Mode, or Web-Only Mode (no keys required).
Installation
This skill requires Python 3 and is typically installed within the Claude Code skills directory.
1. Clone the Skill
Ensure the script and its dependencies are available in your local Claude skills path:
~/.claude/skills/last30days/
2. Configuration (Optional but Recommended)
While the skill works via WebSearch fallback, adding API keys enables engagement metrics (upvotes/likes) which significantly improves the “Judge Agent” synthesis.
mkdir -p ~/.config/last30dayscat > ~/.config/last30days/.env << 'ENVEOF'# last30days API Configuration# Both keys are optional - skill works with WebSearch fallback
# For Reddit research (uses OpenAI's web_search tool)OPENAI_API_KEY=your_openai_key
# For X/Twitter research (uses xAI's x_search tool)XAI_API_KEY=your_xai_keyENVEOF
chmod 600 ~/.config/last30days/.envUsage
You can trigger the skill by describing a topic and an optional target tool.
Argument Syntax
last30days [topic] for [tool] or simply last30days [topic]
Common Examples
- Prompting Research:
last30days photorealistic people for Midjourney - Tool Recommendations:
last30days best Claude Code skills - News & Updates:
last30days latest OpenAI announcements - General Learning:
last30days iOS 19 design trends
How It Works
- Parsing: Claude identifies your Topic, Target Tool, and Query Type.
- Execution: The skill runs a Python script to gather social data and supplements it with a broad
WebSearch. - Synthesis: The internal Judge Agent weights the findings, looking for specific product names, mention counts, and successful prompt structures.
- Summary: You receive a clean summary of findings and engagement stats.
- Tailored Delivery: Once you share your specific vision, Claude generates one perfect prompt in the exact format (JSON, Prose, or Keywords) recommended by the latest research.
Research Depth
You can modify the depth of the research by appending flags to your request:
--quick: Faster execution with 8-12 sources per platform.- (Default): Balanced research with 20-30 sources.
--deep: Comprehensive deep-dive with up to 70 Reddit threads and 60 X posts.
Expert Mode
Once the research phase is complete, Claude enters Expert Mode for that topic. For the remainder of the conversation, the model will:
- Answer follow-up questions using the specific facts, quotes, and insights discovered.
- Cite specific Reddit subreddits or X handles found during the crawl.
- Avoid generic knowledge in favor of the real-time data retrieved.
Note: If no API keys are configured, the skill will automatically fall back to
WebSearchmode. You will still receive high-quality research, though engagement metrics (like upvote counts) will be unavailable.