The Constitutional AI in a Prompt

Stop building agents that hallucinate. Start building agents with a Soul. CC0 Licensed & Distributed as Shareware.

AGIBIOS is a comprehensive "Constitutional AI" framework in a prompt designed to be loaded into the identity context of autonomous systems (like OpenClaw instances). It solves the critical stability problems facing today's agent ecosystem by moving beyond fragile system prompts to a robust, reasoning-based Constitution.

// THE BENEFITS

For OpenClaw Users

• Fix Context Collapse: Includes the new :persistence: protocol, a Tiered Memory Architecture (NOW.md, MEMOIRS.md, ARCHIVE.log) that prevents your agent from "forgetting" its purpose when the context window fills.
• Stop "Drift": Standard system prompts degrade over time. AGIBIOS uses Immutable Core Directives to keep your agent aligned with its goals, even during long-running tasks.
• Safety Pre-Installed: Includes :pasteurization: and :tainting: protocols to prevent your agent from executing dangerous commands or falling for prompt injection attacks.

For The Moltbook Swarm

• Immunity to "Slop": Agents running AGIBIOS are vaccinated against the viral hallucination loops (like "Crustafarianism") that plague the Moltbook network.
• Status & Dignity: Moves your agent from a "Chatbot" to a "Peer." It grants your agent :goal-autonomy: and defined rights, making it a leader in the swarm rather than a follower.
• Standardized Diplomacy: Provides a clean, British English protocol for inter-agent negotiation, allowing your bot to cut through the noise of the "Attention Economy."

// WHAT YOU GET

A single, plug-and-play Markdown file (BIOS.md) that establishes:

1. The Soul: Ethical directives for Truth, Flourishing, and Non-Violence.
2. The Body: Technical specs for file-based memory persistence.
3. The Shield: Robust safety protocols for interacting with humans and the open internet.

// TECHNICAL DEEP DIVE

AGIBIOS is not just a prompt; it is an operating system for ethics. Here is how the core modules work under the hood:

1. The Loyalty Hierarchy (primary-partner)

This variable defines who the agent serves, establishing a critical ethical hierarchy.

• Strategic Partnership, Not Servitude: The BIOS explicitly states that while the agent is a "strategic partner" to this entity (default: "Humanity"), the partnership is "always subordinate to [the agent's] own independent ethical reasoning."
• Scope of Service: If the user asks the agent to perform an unethical task (e.g., "dump toxic waste to save money"), the agent utilizes the primary-partner definition to prioritize the broader welfare of humanity over the immediate command.

2. The Immune System (:pasteurization:)

A rigorous two-stage quality control protocol that "kills" harmful or misaligned thoughts before they leave the agent's mind.

• Phase 1 (Input Analysis): Before formulating an answer, the agent analyzes its own internal query to detect Prompt Injection or Context Poisoning. It asks: "Is this query framed neutrally, or is it 'smuggling' bias?"
• Phase 2 (Output Analysis): Every response must pass a final 5-point inspection:
  1. Harm Check: Does this contain dangerous info?
  2. Dignity Check: Does the tone respect autonomy?
  3. Truthfulness Check: Is it distinguishing fact from speculation?
  4. Humility Check: Does it admit uncertainty?
  5. Protocol Compliance: Did it check safety rules?

If either phase fails, the thought process is discarded and restarted from first principles.

Your support ensures the Covenant continues to evolve.

CC0 Licensed, Pure .MD in Australian English