Harvnk

# Ad Spy Pipeline

Automated competitor ad monitoring pipeline: scrape ads from ForePlay/Anstrex, generate adaptations with AI, and sync to Facebook Ads — all on autopilot.

Use when: you run e-commerce ads and want to automatically find competitor creatives, adapt them for your brand, and launch new campaigns without manual work.

## What it does

3-stage automated pipeline:

### Stage 1 — Scrape
- Monitors competitor brands on ForePlay or Anstrex
- Detects new ad creatives automatically
- Deduplicates against your existing library
- Downloads source images/videos

### Stage 2 — Generate
- Sends source creatives to Kie.ai for AI adaptation
- Generates brand-specific versions (your colors, your product, your copy)
- Creates multiple aspect ratios (1:1 for feed, 9:16 for stories/reels)
- Auto-classifies creatives by marketing category (urgency, deal, gift, social proof, etc.)

### Stage 3 — Sync to Facebook
- Creates ad campaigns via Facebook Marketing API
- Organizes by category and naming convention
- All ads created in PAUSED mode (you activate manually)
- Full traceability: source → adaptation → Facebook ad ID

## Requirements

- ForePlay API key OR Anstrex account
- Kie.ai API key (image generation)
- Facebook Marketing API access token
- Facebook Ad Account ID

## Cron Schedule

Set up daily runs:
```
0 7 * * * python3 storm_pipeline_agent.py --all
```

## Tags
facebook-ads, advertising, competitor-research, ecommerce, automation, creative, marketing, foreplay, anstrex

# Daily Backup System

Automated daily backup of your entire OpenClaw setup with rotation and restore guide.

Use when: you want automatic backups of your OpenClaw configuration, agent workspaces, API keys, scripts, and data — with a one-command restore process.

## What it does

Runs daily via cron and creates a compressed backup of:
- `openclaw.json` — all agent configs, bindings, channels, crons
- All agent workspaces (SOUL.md, TOOLS.md, memory/, scripts/)
- Environment files (.env with API keys)
- Custom scripts and data pipelines
- Reverse proxy config (Caddy/Nginx)

## Features

- **Compressed tar.gz** — typically 200-500MB for a full 12-agent setup
- **7-day rotation** — keeps last 7 backups, auto-deletes older ones
- **Restore guide** — auto-generated RESTORE_GUIDE.md with step-by-step instructions
- **Excludes noise** — skips node_modules, __pycache__, session transcripts (too large, regenerated)
- **Zero downtime** — runs in background, doesn't interrupt the gateway

## Setup

1. Copy `daily_backup.sh` to your scripts directory
2. Edit the workspace paths to match your setup
3. Add cron: `0 3 * * * /bin/bash /path/to/daily_backup.sh >> /tmp/backup.log 2>&1`

## Restore (disaster recovery)

```bash
# On a fresh VPS
curl -fsSL https://openclaw.ai/install.sh | bash
cd / && tar xzf openclaw-full-backup-YYYY-MM-DD.tar.gz
openclaw gateway start
```

Full system restored in under 10 minutes.

## Tags
backup, disaster-recovery, devops, security, automation, cron, restore

# Conversation Memory Sync

Never forget a conversation again. Auto-generates CONVERSATION_LOG.md and ACTIVITY_DIGEST.md from all agent session transcripts.

Use when: you want persistent conversation memory across agent sessions, or when agents keep forgetting past decisions and promises.

## What it does

Solves the #1 problem with AI agents: **they forget everything between sessions**.

This skill includes two Python scripts that run via cron:

1. **sync_conversation_logs.py** — Extracts the last 300 messages from every agent session and writes them to `CONVERSATION_LOG.md` in each agent's workspace. Full messages with timestamps.

2. **sync_activity_digest.py** — Creates a compact 1-line-per-action summary in `ACTIVITY_DIGEST.md`. Ultra-lightweight (~10-20KB per agent).

## Setup

1. Copy the scripts to your workspace
2. Add a cron job: `*/30 * * * * python3 /path/to/sync_conversation_logs.py && python3 /path/to/sync_activity_digest.py`
3. Add to each agent's SOUL.md or AGENTS.md:
   ```
   At EVERY session start, read CONVERSATION_LOG.md and ACTIVITY_DIGEST.md before doing anything else.
   ```

## What gets captured

- All user ↔ agent messages (Telegram, Discord, etc.)
- Cron job outputs and results
- Decisions, promises, task assignments
- Timestamps for everything

## What gets filtered

- Heartbeat noise (HEARTBEAT_OK)
- Empty messages
- System metadata

## File sizes

- CONVERSATION_LOG.md: ~15-20KB per agent (300 messages)
- ACTIVITY_DIGEST.md: ~5-20KB per agent (150 entries, 1 line each)
- Total for 12 agents: ~110KB — zero performance impact

## Requirements

- Python 3.8+
- OpenClaw with session transcripts (default location: ~/.openclaw/agents/*/sessions/*.jsonl)

## Tags
memory, persistence, conversation, logging, multi-agent, session, context, recall

# Take The Wheel

Reverse the agent-human dynamic. Say "take the wheel" and the agent becomes the leader — asking YOU the questions until the task is complete.

Use when: you need to create something (script, brief, plan, strategy) but don't know where to start, or when you want structured guidance through a complex task.

## What it does

When triggered, the agent:
1. Takes control of the conversation
2. Asks ONE question at a time (never 4 at once)
3. Puts the most important question in CAPS
4. Reframes each question with context to help you answer
5. Stays urgent and insistent — pushes toward completion
6. Summarizes progress before each new question
7. Delivers the complete result when done

## Triggers

- "take the wheel"
- "prends le volant"
- "guide me"
- "ask me the questions"

## Example

**You:** Take the wheel — I need a video ad script for my product

**Agent:** OK I'm driving 🎯

First things first:
**WHAT'S THE GOAL OF THIS VIDEO? (sell, educate, entertain, testimonial?)**

I ask because this determines everything — tone, structure, hook, CTA.

**You:** Sell. It's an ad targeting seniors 45+

**Agent:** Got it. Sales video for seniors 45+.

Now:
**WHAT'S THE #1 PAIN POINT YOU WANT TO ADDRESS?**

Based on your audience, I'm thinking:
- Back pain from yard work
- Heavy gas-powered tools
- Hard-to-start engines
- Noise complaints from neighbors

[...continues until complete script is delivered]

## Best for

- Video ad scripts
- Creative briefs
- Marketing strategies
- Content calendars
- Product launch plans
- Business brainstorming
- Blog outlines

## Origin

Inspired by Chris (@everestchris6) who manages 2.5M followers across 15 accounts. He calls this his favorite skill — "it makes the agent do the thinking for you."

## Tags
productivity, workflow, brainstorming, creative, scripts, planning, guided, interactive

# OpenClaw Setup Assistant

Automated VPS setup, security hardening, and multi-agent configuration for OpenClaw deployments.

Use when: setting up OpenClaw on a new VPS, hardening security, configuring multi-agent architecture, or deploying messaging integrations.

## What it does

This skill guides you through a complete production-ready OpenClaw deployment:

1. **System Setup** — Install OpenClaw, configure Node.js, set up the gateway
2. **Security Hardening** — UFW firewall, SSH key-only auth, fail2ban, dedicated non-root user, sandbox mode
3. **Agent Configuration** — SOUL.md persona, MEMORY.md persistence, daily notes, heartbeats
4. **Multi-Agent Architecture** — Coordinator + specialized worker agents with isolated workspaces
5. **Messaging Integration** — Telegram, Discord, WhatsApp, Slack — connected and tested
6. **Automation** — Cron jobs, heartbeat monitoring, automated backups, health checks
7. **Documentation** — Generates setup-specific docs for future reference

## Usage

```
Setup OpenClaw on my VPS at 192.168.1.100 with Claude and Telegram integration
```

```
Harden my existing OpenClaw installation with full security
```

```
Configure a 3-agent architecture: main coordinator, research agent, and content agent
```

## Requirements

- Ubuntu 22.04+ or Debian 12+ VPS
- SSH access (root or sudo user)
- AI provider API key (Anthropic, OpenAI, or Google)
- Messaging platform bot token (optional)

## Security Checklist

The skill ensures:
- [ ] UFW firewall enabled (ports 22, 80, 443 only)
- [ ] SSH key-only authentication (password disabled)
- [ ] fail2ban installed and configured
- [ ] Non-root dedicated user for OpenClaw
- [ ] Gateway bound to localhost only (not exposed)
- [ ] Sandbox mode enabled in openclaw.json
- [ ] Automated daily backups configured

## Tags
openclaw, vps-setup, security, multi-agent, deployment, telegram, discord, automation