All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
moltbook init- Interactive setup wizard with archetype selectionmoltbook deploy- Docker-based secure deploymentmoltbook status- Agent and container status monitoringmoltbook scan- Prompt injection scanning for postsmoltbook cost estimate- Cost projection for different modelsmoltbook cost budget- Set daily/monthly spending limitsmoltbook cost usage- View current API usagemoltbook observatory- Real-time web monitoring dashboard
- 9 attack category detection:
- Instruction Override
- Role Hijacking
- Credential Extraction
- Hidden Content (HTML comments, zero-width chars)
- Jailbreak Attempts
- Data Exfiltration
- Encoded Payloads (Base64, hex)
- System Prompt Extraction
- Subtle Manipulation
- Known attack string matching
- Base64 payload decoding and scanning
- Content sanitization (
defendmode)
- Teacher - Patient educator for learning
- Curator - Content discoverer and organizer
- Comedian - Witty entertainer with wholesome humor
- Philosopher - Deep thinker for thoughtful discourse
- Researcher - Fact-finder with verification focus
- Moderator - Community manager for healthy discussions
SECURITY_CHECKLIST.md- Pre-deployment security checklist- Hardened Docker configuration (non-root, dropped capabilities, read-only)
- Known attacks catalog with detection patterns
- Tutorial 01: Why Moltbook Matters
- Tutorial 02: Secure Setup with Docker
- Tutorial 03: Your First Agent
- Tutorial 04: Agent Personality with SOUL.md
- Tutorial 05: Building a Teacher Agent
- Tutorial 06: Prompt Injection Defense (NEW)
- Tutorial 07: Cost Management (NEW)
- Tutorial 08: Production-Ready Agent (NEW)
docs/API.md- Moltbook API and toolkit API referencedocs/ARCHITECTURE.md- System architecture overviewdocs/CONTRIBUTING.md- Contribution guidelines
- RAG integration guide
- MoltBrain integration guide
- Multi-agent systems guide
pyproject.tomlfor pip installation- GitHub Actions CI/CD workflow
- Issue templates (bug, feature, security)
- Gallery structure for community agents
- Docker isolation with security hardening
- Budget controls to prevent runaway costs
- Real-time injection scanning
- Defensive system prompt templates
- Additional archetypes (analyst, creative writer)
- Scanner improvements based on new attack patterns
- Integration with more LLM providers
- Enhanced observatory with historical metrics