last72hours

---
name: last72hours
description: Use when you want a 72-hour viral radar across Reddit, X, TikTok, Instagram, HN, YouTube, and GitHub — clickable HTML mirror plus an optional editorial leaderboard rendered in Paper.design.
source: https://github.com/iharnoor/last72hours-skill
generated: 2026-05-12T18:04:53.770Z
category: tool
audience: creators
---

## Tutorials

- https://skillmake.xyz/v/last72hours.mp4

## When to use

- Spotting what's spiking right now (last 72 hours) on a launch, news beat, or competitor across 7 platforms in one pass
- Producing a clickable HTML brief with source links and engagement counts, instead of manually sweeping Reddit/X/TikTok/HN
- Generating an editorial-scorecard leaderboard artboard in Paper.design straight from the radar output (via Paper MCP)
- Running a free-only Reddit + X subscan when you don't want to burn ScrapeCreators credits on IG/TikTok

## Key concepts

### 72-hour viral radar

Scrapes the past 72 hours of activity across Reddit, X, TikTok, Instagram, Hacker News, YouTube, and GitHub (plus optional Bluesky) for a given topic. Each item carries URL, score, and quoted snippet — the synthesis stays grounded in real posts, not training data.

### Self-contained HTML mirror

Output is a single static HTML page with no external dependencies — clickable source links, embedded counts, no JS framework. Lives under `LAST72_OUTPUT_DIR` (defaults to `~/Documents/Last72Hours`) so the brief survives the agent session.

### Paper.design artboard via MCP

Optional: if Paper Desktop is running and its MCP server is attached, the skill builds an editorial-scorecard leaderboard artboard from the same radar data and exports a 2× PNG preview. Auto-skips cleanly when Paper isn't attached.

### Free vs paid sources

Free with browser cookies: Reddit, X (via AUTH_TOKEN/CT0), Hacker News, YouTube, GitHub. Paid via ScrapeCreators: Instagram and TikTok. The bundled `last72-reddit-x` subagent runs a Reddit + X-only sweep when you want to avoid SCRAPECREATORS_API_KEY charges entirely.

### Built on last30days

Sibling/downstream of [[last30days]] — same evidence-grounded synthesis pattern, same engine plumbing, narrower time window. Install both; they coexist as separate slash commands.

## API reference

```
Install the dependency (last30days-skill)
```

last72hours layers on top of the last30days engine — install that first.

```
git clone https://github.com/mvanhorn/last30days-skill ~/.last30days-skill
ln -s ~/.last30days-skill/skills/last30days ~/.claude/skills/last30days
```

```
Install last72hours
```

Clone the repo and symlink the skill folder into Claude Code's user skills directory.

```
git clone https://github.com/iharnoor/last72hours-skill ~/.last72hours-skill
ln -s ~/.last72hours-skill/skills/last72hours ~/.claude/skills/last72hours
```

```
Optional: attach Paper Desktop MCP
```

Enables the editorial-scorecard leaderboard artboard step. Skip if you don't have Paper Desktop running.

```
claude mcp add paper --transport http http://127.0.0.1:29979/mcp --scope user
```

```
/last72hours <topic>
```

User-invocable slash command. Triggers the full radar-and-synthesize pipeline. Output lands in `LAST72_OUTPUT_DIR` (defaults to `~/Documents/Last72Hours`).

```
/last72hours AI coding agents
/last72hours nvidia earnings
/last72hours claude code skills
```

```
Reddit + X-only subagent
```

Run a fast, free pulse without burning ScrapeCreators credits on IG/TikTok. Useful for dev-channel-heavy topics.

```
# Invoked from inside the host agent via the Task tool —
# pick the `last72-reddit-x` subagent shipped at agents/last72-reddit-x.md
# (no IG/TikTok scraping, no SCRAPECREATORS_API_KEY required)
```

```
Required & optional env vars
```

Reddit, HN, YouTube, GitHub are free with no keys. X needs either browser cookies (free) or XAI_API_KEY (paid). Instagram + TikTok need SCRAPECREATORS_API_KEY.

```
SCRAPECREATORS_API_KEY=...   # Instagram, TikTok (paid)
AUTH_TOKEN=...               # X auth cookie (free, from browser)
CT0=...                      # X CSRF cookie (free, from browser)
XAI_API_KEY=...              # Alternative X API (paid)
BSKY_HANDLE=...              # Optional Bluesky
BSKY_APP_PASSWORD=...
LAST72_OUTPUT_DIR=~/Documents/Last72Hours
```

## Gotchas

- Requires the upstream `last30days-skill` to be installed first — it provides the shared engine plumbing this skill calls into.
- Instagram and TikTok scraping costs ScrapeCreators credits. If you only need a pulse, use the `last72-reddit-x` subagent — it skips both platforms entirely.
- X scraping defaults to browser cookies (AUTH_TOKEN + CT0). Cookies expire — refresh them when X results suddenly drop to zero.
- Paper.design artboard generation is opt-in: it runs only when Paper Desktop's MCP server is attached. Without it the skill still produces the HTML mirror — it just skips the visual leaderboard step.
- Output is deterministic per topic + window — re-running on the same topic within the 72h window will overwrite, not append. Move or rename the HTML if you want to keep the prior pass.
- Time window is fixed at 72 hours — for longer recency, switch to [[last30days]] rather than trying to widen this skill.
- Self-contained HTML means no external CDN calls — it works offline, but also means no live engagement counters. Numbers are frozen at scan time.

---
Generated by SkillMake from https://github.com/iharnoor/last72hours-skill on 2026-05-12T18:04:53.770Z.
Verify against source before relying on details.

File: ~/.claude/skills/last72hours/SKILL.md