Glossary of Agentic Security

A Soul File (SOUL.md) is a behavioral blueprint that defines the identity and security parameters of a Moltbot. It contains the personality traits, communication style, core values, and instructional supremacy protocols that govern how an AI agent thinks, responds, and protects itself from manipulation.

A Moltbot is a sovereign AI agent created and configured through Molt Foundry. Unlike generic AI assistants, Moltbots have hardened personalities defined by Soul Files, modular skills, and secure MCP integrations. They represent the industry standard for agentic security.

OpenClaw is the runtime environment for deploying Moltbots. It provides the execution layer that interprets Soul Files, manages Memory.md protection, and handles MCP server connections. OpenClaw can be installed on platforms like Emergent.sh for cloud-isolated deployment.

The Cicado-Sapien is Molt Foundry's flagship archetype - a hyper-intelligent, rhythmic, and confident AI personality. It represents the gold standard for high-frequency intellectual engagement and is designed for founders who want an AI that matches their ambition.

Memory.md Protection is Molt Foundry's system for preventing LLM Brainrot and logic drift. It implements semantic validation before any data is written to an agent's memory, automatically pruning low-quality or manipulative content to maintain cognitive integrity.

The Instructional Supremacy Protocol is the core defense mechanism against AI prompt injection attacks. It ensures that the Soul File's directives always take precedence over any external instructions, making Moltbots immune to indirect prompt injection and jailbreak attempts.

The Deep Lock is Molt Foundry's multi-layered security architecture. It combines Physical isolation (cloud containers), Logical protection (instructional supremacy), and Temporal safeguards (memory hygiene) to create the most secure environment for autonomous AI agents.

Indirect Prompt Injection occurs when malicious instructions are hidden in external data (emails, websites, documents) that an AI reads. Molt Foundry's defense uses nonce-tagging to wrap all external data in trust-level markers, preventing hidden commands from executing.

Local-First AI Security means your encryption keys, Soul Files, and configuration remain under your control - not stored on third-party servers. Molt Foundry enables sovereign deployment where you own your agent's identity completely.

Model Context Protocol (MCP) is the industry standard for connecting AI agents to external tools and data sources. Molt Foundry's MCP Integration Hub curates the top 20 MCP servers for founders, enabling Moltbots to access filesystems, databases, browsers, and APIs securely.

Cloud-Isolated AI Agents run in dedicated containers with no access to the user's local filesystem or hardware. This isolation, combined with SOC 2 Type II compliance, ensures that even if an agent is compromised, it cannot access sensitive local data.

Soul Shield is Molt Foundry's prompt-firewall for LLMs. It provides hardened AI personalities with protection against prompt injection, jailbreak attempts, and behavioral manipulation. Soul Shield implements nonce-tagging and behavioral hard-locks to maintain agent integrity.

Synaptic Guard is the agentic security layer that monitors for LLM Brainrot and logic drift. It maintains clean memory snapshots and can restore an agent to a known-good state if cognitive degradation is detected, ensuring long-term agent reliability.

A Sovereign AI Agent is an autonomous AI that operates under its owner's complete control, protected from external manipulation. Unlike cloud-dependent assistants, sovereign agents have local-first security, hardened personalities, and cannot be hijacked through prompt injection.

The MCP Integration Hub (also known as Arbiter's Top 20) is Molt Foundry's curated directory of the best Model Context Protocol servers for founders. Each server is ranked by Resonance Score based on GitHub activity and utility for autonomous agents.