tools-only last30days - Community Research Skill
Research what people are saying about a topic across Reddit, X, Hacker News, and developer communities from the last 30 days.
Description
Triggers on: "what's trending", "what are people saying about", "research", "last 30 days", "community sentiment", "what's new with", "recent discussions"
This skill searches multiple community platforms to surface real conversations, opinions, and sentiment about any topic. It combines engagement-weighted social signals with technical community discussions.
Source Tiers
Tier 1: High-Signal (Engagement-Weighted)
- Reddit — Via OpenAI API with web search (requires OPENAI_API_KEY)
- X/Twitter — Via xAI Grok API (requires XAI_API_KEY)
- Hacker News — Free Algolia API, always available
Tier 2: Developer Communities
- Stack Overflow — Free API, technical Q&A
- Dev.to — Free API, developer blog posts
- Lobsters — Free, curated tech discussions
Tier 3: Web Fallback
- General web search — Via web_fetch tool
Configuration
Credentials stored at: ~/.config/last30days/.env
# Optional - enables Reddit search
OPENAI_API_KEY=sk-...
# Optional - enables X/Twitter search
XAI_API_KEY=xai-...
# HN, Stack Overflow, Dev.to, Lobsters need NO API keys
Research Flow
Step 1: Parse Intent
Extract from user query:
- Topic: Main subject to research
- Query type: general sentiment | tool comparison | "how to" | trending | prompting
- Time range: Default 30 days, can be adjusted
Step 2: Search Tier 1 Sources (Parallel)
Run these in parallel using exec tool:
# Reddit + X (if API keys available)
cd ~/clawd-nuri-internal/skills/last30days
python3 scripts/last30days.py "TOPIC" --emit=json
# Hacker News (always available)
python3 scripts/hn_search.py "TOPIC" --days 30 --limit 30
Step 3: Search Tier 2 Sources
cd ~/clawd-nuri-internal/skills/last30days
./scripts/community_search.sh "TOPIC" all
Step 4: Web Fallback (if needed)
Use web_fetch for additional sources if Tier 1/2 results are sparse:
- Blog posts
- News articles
- Documentation
- Tutorial sites
Step 5: Synthesize Results
Weight sources by engagement quality:
| Source | Weight | Reasoning |
|---|---|---|
| Reddit (high upvotes) | 1.0 | Strong community validation |
| X (high engagement) | 0.9 | Real-time pulse |
| Hacker News | 0.85 | Tech-savvy audience |
| Stack Overflow | 0.7 | Technical depth |
| Dev.to | 0.6 | Developer perspective |
| Lobsters | 0.6 | Curated tech |
| Web (general) | 0.4 | No engagement signal |
Step 6: Present Findings
Structure output as:
## Research: [TOPIC] (Last 30 Days)
### Key Themes
1. [Theme with source citations]
2. [Theme with source citations]
### Sentiment Summary
- Overall: [Positive/Neutral/Negative/Mixed]
- Common praise: [...]
- Common criticism: [...]
### Top Discussions
**Reddit** (X posts, Y total upvotes)
- [Title](url) — X upvotes, Y comments — key insight
**X/Twitter** (X posts)
- [Key tweet summary](url) — engagement stats
**Hacker News** (X posts, Y total points)
- [Title](url) — X points, Y comments — key insight
**Stack Overflow** (X questions)
- [Common problem pattern]
### Emerging Patterns
- [Pattern 1]
- [Pattern 2]
Step 7: Generate Prompting Query (If Applicable)
If user asked about prompting/techniques, generate a copy-paste prompt:
## Suggested Prompt (Copy-Paste Ready)
Based on community insights, here's an optimized prompt for [TOPIC]:
---
[Generated prompt incorporating community best practices]
---
Example Usage
User: "What are people saying about Cursor IDE in the last 30 days?"
Flow:
- Parse: topic="Cursor IDE", type="sentiment/opinions"
- Run Reddit/X search (if keys available)
- Run HN search:
python3 scripts/hn_search.py "Cursor IDE" --days 30 - Run community search:
./scripts/community_search.sh "Cursor IDE" all - Synthesize across sources
- Present with engagement stats
User: "Research Claude vs GPT-4 for coding"
Flow:
- Parse: topic="Claude vs GPT-4 coding", type="comparison"
- Search all sources
- Weight comparative discussions higher
- Present pros/cons from each platform
User: "What's the best way to prompt for code review?"
Flow:
- Parse: topic="code review prompts", type="prompting/how-to"
- Search all sources
- Extract specific techniques mentioned
- Generate optimized prompt from community insights
Script Reference
last30days.py (Reddit + X)
python3 scripts/last30days.py "topic" [options]
Options:
--mock Use fixtures (testing)
--emit=MODE compact|json|md|context|path
--sources=MODE auto|reddit|x|both
--quick Fewer results, faster
--deep More comprehensive
--include-web Add web search
hn_search.py (Hacker News)
python3 scripts/hn_search.py "topic" [options]
Options:
--days N Days to look back (default: 30)
--limit N Max results (default: 50)
community_search.sh (SO, Dev.to, Lobsters)
./scripts/community_search.sh "topic" [source]
Sources: stackoverflow, devto, lobsters, all
Error Handling
- No API keys: Fall back to HN + Tier 2 sources (still useful!)
- API errors: Log error, continue with available sources
- No results: Suggest broader topic or different time range
- Rate limits: Wait and retry, or use cached results
Notes
- HN, SO, Dev.to, Lobsters are FREE and always available
- Even without OpenAI/xAI keys, this skill provides valuable research
- Reddit/X add engagement weighting that improves signal quality
- Always cite sources with links for user verification