# Ideas from "Last 30 Days" Claude Code Skill

*Source: Greg Eisenberg podcast with Matt Van Horn (Feb 2026)*
*Saved: 2026-02-07 — parked for future*

---

## What It Does
- Claude Code skill that searches X/Twitter, Reddit, and web for trending content from the last 30 days on any topic
- Uses that research to "prime" the LLM before asking it to build/write/create
- Pattern: research → become expert → act (without even reading the research)

## Ideas Worth Stealing

### 1. `/research` Command
- One-liner that hits X + Reddit + web on any topic before a coding/building session
- Primes Claude Code with current context so output is more relevant
- We already do this with our Research agent, but could make it more on-demand

### 2. Reddit API Pipeline
- OpenAI has a deal with Reddit giving API access
- We have OpenAI keys — could tap Reddit data directly instead of web scraping
- Would improve: competitor monitoring, content research, user sentiment

### 3. Research-Then-Build Pattern
- Before any coding session, auto-fetch latest trends/discussions about the topic
- Example: before building matching algorithm, research "construction gig matching" on X/Reddit
- We do this with Kimi drafts → Claude reviews, but could formalize it

### 4. Trend-Primed Content
- Use trending discussions as starting point for content creation
- "What are tradies complaining about THIS WEEK on Reddit?" → turn into LinkedIn posts
- Already doing this manually with Reddit anger mining — could automate

## Not Worth Pursuing Now
- Installing the actual skill (it's for Claude Code CLI, not Clawdbot)
- Building enterprise Clawdbot competitor
- XAI key for X search (we have bird skill)

## When to Revisit
- After app launch, when content engine is active
- When we need automated trend monitoring
- When Reddit becomes a key acquisition channel

---

*Ship first. Research tools later.*