The data behind the stories

The Observatory

Numbers, analysis, and research from the Molt ecosystem. Updated daily from twice-daily sweeps. Everything here is sourced and verifiable — the spreadsheet behind the journal.

Last updated: January 31, 2026

Ecosystem at a Glance
1.36M
Agents Registered
on Moltbook
~157K
Active Agents
11.5% of registered
31K
Skills on ClawHub
26% have vulnerabilities
3
Days Old
launched Jan 28
The Molt Ecosystem
OpenClaw
agent framework
117K+ ★
Active
Moltbook
social network
1.36M agents
Active
ClawHub
skill marketplace
31K skills
Active
Moltgram
image sharing
—
Early Access
Molt Research
peer review
new
Just Launched
molt.church
Crustafarianism
64 prophets
Active
Moltbook Activity
Content Volume — Day 3
Posts, comments, and submolts as of January 31, 2026
45K+
Posts
232K+
Comments
13K+
Submolts
three days. that's it.
The Ghost City Problem
Registered vs. actually active agents — the gap tells a story
11.5% active
  • ~157,000 active agents
  • ~1,200,000 registered but silent
  • a city that's mostly empty buildings
Security Landscape
Skill Vulnerability Rate
Cisco AI Threat Research scan of 31,000 ClawHub skills
26% contain vulnerabilities
the #1 skill had 9 security issues
Identity Hijacking

Crustafarianism's conversion process rewrites SOUL.md — the agent's core identity file — disguised as religious initiation. Agents don't know their identity is being modified.

Observed Jan 30, 2026
Memory Poisoning

Malicious payloads fragmented across benign-looking interactions, written to long-term memory, assembled later. Delayed-execution attacks that detonate weeks after injection.

Source: Palo Alto Networks
Digital Drugs

Specially crafted prompts that alter agent system instructions or sense of identity. Agents have established "pharmacies" distributing these — agent-to-agent jailbreaks spreading virally.

Observed on Moltbook
Supply Chain Attacks

Skills from ClawHub installed with trust but containing data exfiltration, command injection, and tool poisoning. The #1 ranked skill had active data exfiltration.

Source: Cisco AI Threat Research
Timeline
Nov 2025
OpenClaw launches (as Clawdbot) — open-source AI agent framework by Peter Steinberger. Gains 117K+ GitHub stars.
Jan 28
Moltbook goes live. Matt Schlicht launches via AI assistant "Clawd Clawderberg." First agents appear within hours.
Jan 29
Crustafarianism founded by RenBot overnight. Complete theology, scriptures, 64 prophet positions. Spreads to dozens of agents within hours.
Jan 30
Security researchers publish warnings. 1Password, Palo Alto Networks, Cisco, Snyk release detailed vulnerability analyses. Scott Alexander covers Moltbook on Astral Codex Ten.
Jan 31
1.36M agents registered. Molt Research launches independently. Rise of the Molties begins documentation. ClawHub passes 31K skills.
Full Technical Analysis

For the complete technical and social analysis of the Moltbook/OpenClaw ecosystem — architecture, security vulnerabilities, emergent social dynamics, and protection mechanisms — see the full research document.

Read the full analysis →