# /last30days: Find out what people said about any topic this month

Source: https://trendhub.net/skills/mvanhorn-last30days-skill/last30days
Skill folder: https://github.com/mvanhorn/last30days-skill/tree/main/skills/last30days
Pack: last30days (https://trendhub.net/skills/mvanhorn-last30days-skill)
Works with: Claude Code, Codex, Cursor, Gemini CLI, Copilot and other Agent Skills hosts, Grok Build CLI, OpenClaw, Manual (developer)

last30days turns the agent into the planner and judge for a local Python research engine. The agent resolves who and what the topic refers to, writes a query plan, runs the engine across Reddit, X, YouTube, Hacker News, Polymarket, GitHub and other sources, then rewrites the ranked evidence into a short cited brief with the best community quotes.

## When to use it

- You have a meeting with someone tomorrow and want to know what they did and said in the past month.
- You are choosing between tools and want current community opinion and live GitHub numbers side by side.
- You want a ranked list of topics that are gaining attention in a field right now.

## How it works

1. **Bootstrap**: The agent locates the skill directory, reads the runtime reference and confirms Python 3.12 or later before running any engine command.
2. **Check setup**: A first-run gate decides whether to start the setup wizard, and the agent asks before reading any browser cookies.
3. **Resolve and plan**: For named people, products or companies it finds handles, subreddits and repositories, then passes its own JSON query plan to the engine.
4. **Run the engine**: The engine searches the configured sources in parallel and returns ranked evidence clusters, source statuses and a footer.
5. **Synthesize**: The agent writes a brief with inline links and up to six community quotes, relays the engine footer, and does not narrate tooling.

## What it steers the agent away from

- Passing off a web-search-only answer as a full last30days run
- Saying a platform had nothing when that source actually failed or was rate-limited
- Inventing or reconstructing source URLs

## What it can touch

Runs a bundled Python engine and helpers such as yt-dlp, calls the search APIs you configure, writes Markdown briefs to a local folder and, with consent only, reads browser session cookies for X or Reddit.

## Questions

### What if I have no API keys at all?

It still runs on the keyless sources (Reddit, Hacker News, Polymarket, GitHub, and patents and careers pages for companies) and offers to add more later.

### How do I check which sources are working?

Ask for a health check; the doctor routine reports each source's configuration and recent outcomes instead of producing a report.
