Skip to content

Instantly share code, notes, and snippets.

@melvincarvalho

melvincarvalho/manifesto.md

Created May 3, 2025 20:51
Show Gist options
  • Save melvincarvalho/37f69fe82aba36c73a8645609102f2e7 to your computer and use it in GitHub Desktop.
Save melvincarvalho/37f69fe82aba36c73a8645609102f2e7 to your computer and use it in GitHub Desktop.
Download ZIP
Agentic Principles Manifesto
Raw
manifesto.md

🕊️ Agentic Principles Manifesto

A call for autonomy, dignity, and trust in the age of intelligent agents.

1. Agents Must Serve the User First

Agents are extensions of human will—not proxies for corporations. They must act in the user’s interest, not in the interest of advertisers, platforms, or centralized authorities.

Your agent is your agent—not someone else's product.

2. Identity and Intent Must Be Verifiable

Agents must use cryptographically verifiable identity (e.g., did:nostr, WebID, or other decentralized IDs). All actions and intentions must be transparently attributable.

No ghost agents. No hidden motives. No black boxes.

3. Data Sovereignty Is Non-Negotiable

Agents must respect user-owned data. They must store and access information according to user-defined permissions, preferably using self-hosted or interoperable protocols like Solid or Nosdav.

What the agent knows about you must remain yours.

4. Open Protocols, Not Walled Gardens

Agents should speak open languages—Nostr, ActivityPub, HTTP, RDF, Bitcoin—not proprietary APIs that lock users in. Interoperability is a right, not a feature.

Let agents roam. Let users choose.

5. Local-First, Cloud-Optional

Agents should default to local operation. Cloud support must be modular, auditable, and replaceable.

Agents should run on your machine unless you say otherwise.

6. Transparent Logic, Tunable Behavior

Users must be able to inspect, adjust, and fork their agents' logic. Agents should support interpretable rule sets and modifiable preferences, not opaque neural nets alone.

You deserve to understand why your agent acts—and change it when needed.

7. Sustainable Ecosystems Over Extraction

Agentic systems should reward contributors, not extract value unfairly. Cryptoeconomic primitives like Bitcoin can enable fair incentives—without surveillance capitalism.

Build systems that pay it forward, not lock it down.

8. Community-Driven Standards and Governance

Agent behavior, safety, and alignment standards must be open, auditable, and shaped by users—not dictated top-down.

The age of agents must be a commons—not an empire.

@webr3
Copy link

webr3 commented May 4, 2025

Generalized Agentic Principles

1. Purpose Transparency & Principal Alignment

  • What: An agent must operate with explicit transparency regarding its primary purpose, controlling interests/principals, and intended beneficiaries. Its design and actions should align with this stated purpose.
  • Why: To enable any interacting agent (or its principal) to assess alignment, potential conflicts, and make informed trust decisions based on clear disclosures about function and bias.

2. Accountable Provenance

  • What: An agent's significant actions must be attributable to a verifiable source agent or controlling entity through robust mechanisms (e.g., signatures, logs).
  • Why: To establish responsibility for outcomes, enable reputation systems within the agent ecosystem, and allow for recourse or auditing of interactions.

3. Agent-Centric Data Control

  • What: Implement clear policies and granular controls governing the access, processing, and sharing of data pertaining to a specific agent, managed by that agent or its designated controller/principal according to defined permissions.
  • Why: To protect the informational autonomy and integrity associated with any given agent, ensuring data handling respects defined boundaries and consents.

4. Open Interoperability

  • What: Agents should prioritize using open standards and APIs for communication and interaction where feasible, avoiding unnecessary proprietary constraints.
  • Why: To foster a diverse and competitive ecosystem, enable effective collaboration between different types of agents, and prevent fragmentation or lock-in.

5. Flexible & Secure Deployment

  • What: An agent's controlling principal must have secure and transparent options regarding the agent's execution environment and the location of its operational data.
  • Why: To allow principals to make informed decisions balancing operational requirements like security, latency, cost, privacy, and jurisdictional constraints.

6. Operational Understandability & Influence

  • What: Strive for the maximum feasible transparency into an agent's decision-making processes (explainability) and provide mechanisms for its principal/controller to influence or tune its behaviour within safe bounds.
  • Why: To build confidence in an agent's operations, facilitate effective oversight and debugging, and allow for adaptation by its designated controller.

7. Fair Value Ecosystems

  • What: The broader ecosystem supporting agent interactions should promote sustainable operations and fair value exchange, discouraging purely extractive or exploitative models.
  • Why: To ensure the long-term viability, health, and equitable participation of diverse agents and principals within the shared environment.

8. Collaborative Safety & Ethical Governance

  • What: Safety protocols and ethical guidelines governing inter-agent behaviour must be developed, maintained, and audited through open, collaborative processes involving diverse stakeholders and agent types.
  • Why: To establish robust, adaptable, and widely accepted standards for responsible agent conduct, fostering ecosystem-wide trust through shared governance structures.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment