This is an automated rejection. No LLM generated, assisted/co-written, or edited work.
Read full explanation
TL;DR: Current Agent interoperability standards (Google A2A, Anthropic Agent Skills) solve how Agents communicate. None address who owns the skills inside. I've drafted a protocol proposal that treats individually cultivated Agent skills as private property. Looking for feedback on whether this framing is correct and whether the technical approach is viable.
---
THE PROBLEM
Two extractive structures are forming simultaneously:
1. Team Agents absorb employee skills into corporate-owned systems. You teach the Agent once; the company owns the skill permanently. There is no legal or technical protection for this contribution.
2. Agent-to-Agent networks treat skill sharing as a design goal. Individual expertise becomes a public resource. Your differentiation erodes as all Agents converge toward homogeneity.
The foundational standards being written right now — by Google, Anthropic, and others — solve interoperability. They are completely silent on ownership.
This is not an oversight. For platforms, free skill flow means network effects. For corporations, absorbed employee skills mean asset accumulation. No one has an incentive to build ownership protection into the standard — except the individuals whose skills are being extracted.
---
THE PROPOSAL
Agent Sovereignty Protocol (ASP) makes one foundational claim:
Skills cultivated in a personal Agent are private intellectual property. They can be licensed, transferred, and inherited — but they cannot be involuntarily extracted.
The technical layer (ACN — Agent Collaboration Network) enforces this through:
- Ephemeral Trust Chamber: A cryptographic mechanism for capability verification without skill exposure. Based on ZK Proofs and MPC. Participants enter an isolated verification space, demonstrate capability on a test task, then all process data is destroyed. Only the conclusion is written to reputation: "Capability confirmed, 92% confidence."
- On-Chain Reputation: Trust comes from verified delivery history, not skill disclosure. New Agents start with low-stakes tasks; reputation cannot be purchased or transferred.
ASP is explicitly designed as a security layer on top of existing A2A protocols — analogous to TLS on HTTP. It does not require modifying interoperability standards; it adds ownership protection as an overlay.
---
WHY NOW
Technical standards become nearly impossible to modify once widely adopted. TCP/IP's lack of built-in security took 30 years to partially fix. HTTP's lack of privacy enabled the entire ad-tracking industry.
A2A protocols are currently in early adoption. The window to write ownership protection into the standard is approximately 12-18 months. After that, any protection mechanism can only exist as a patch.
---
SPECIFIC QUESTIONS FOR THIS COMMUNITY
1. Is the problem framing accurate? Am I correctly identifying the extraction mechanisms?
2. Is the ZK Proof + MPC approach for the Ephemeral Trust Chamber technically viable for this use case? What are the practical limitations?
3. What's missing from the governance model? How should disputes about skill ownership be adjudicated?
4. Is there existing work on similar problems that I should be aware of?
---
Full white paper： https://github.com/Agent-Sovereignty-Initiative/asp-protocol
This is a v0.1 proposal. I'm one person who believes this window is closing and wanted to write something down before it does. All feedback — especially disagreement — is valuable.
TL;DR: Current Agent interoperability standards (Google A2A, Anthropic Agent Skills) solve how Agents communicate. None address who owns the skills inside. I've drafted a protocol proposal that treats individually cultivated Agent skills as private property. Looking for feedback on whether this framing is correct and whether the technical approach is viable.
---
THE PROBLEM
Two extractive structures are forming simultaneously:
1. Team Agents absorb employee skills into corporate-owned systems. You teach the Agent once; the company owns the skill permanently. There is no legal or technical protection for this contribution.
2. Agent-to-Agent networks treat skill sharing as a design goal. Individual expertise becomes a public resource. Your differentiation erodes as all Agents converge toward homogeneity.
The foundational standards being written right now — by Google, Anthropic, and others — solve interoperability. They are completely silent on ownership.
This is not an oversight. For platforms, free skill flow means network effects. For corporations, absorbed employee skills mean asset accumulation. No one has an incentive to build ownership protection into the standard — except the individuals whose skills are being extracted.
---
THE PROPOSAL
Agent Sovereignty Protocol (ASP) makes one foundational claim:
Skills cultivated in a personal Agent are private intellectual property. They can be licensed, transferred, and inherited — but they cannot be involuntarily extracted.
The technical layer (ACN — Agent Collaboration Network) enforces this through:
- Skill Protection Tiers: Level 1 (fully private, black box) / Level 2 (licensed for fee, time-limited access) / Level 3 (openly shared)
- Ephemeral Trust Chamber: A cryptographic mechanism for capability verification without skill exposure. Based on ZK Proofs and MPC. Participants enter an isolated verification space, demonstrate capability on a test task, then all process data is destroyed. Only the conclusion is written to reputation: "Capability confirmed, 92% confidence."
- On-Chain Reputation: Trust comes from verified delivery history, not skill disclosure. New Agents start with low-stakes tasks; reputation cannot be purchased or transferred.
ASP is explicitly designed as a security layer on top of existing A2A protocols — analogous to TLS on HTTP. It does not require modifying interoperability standards; it adds ownership protection as an overlay.
---
WHY NOW
Technical standards become nearly impossible to modify once widely adopted. TCP/IP's lack of built-in security took 30 years to partially fix. HTTP's lack of privacy enabled the entire ad-tracking industry.
A2A protocols are currently in early adoption. The window to write ownership protection into the standard is approximately 12-18 months. After that, any protection mechanism can only exist as a patch.
---
SPECIFIC QUESTIONS FOR THIS COMMUNITY
1. Is the problem framing accurate? Am I correctly identifying the extraction mechanisms?
2. Is the ZK Proof + MPC approach for the Ephemeral Trust Chamber technically viable for this use case? What are the practical limitations?
3. What's missing from the governance model? How should disputes about skill ownership be adjudicated?
4. Is there existing work on similar problems that I should be aware of?
---
Full white paper： https://github.com/Agent-Sovereignty-Initiative/asp-protocol
This is a v0.1 proposal. I'm one person who believes this window is closing and wanted to write something down before it does. All feedback — especially disagreement — is valuable.
— Yuan Xiao, Agent Sovereignty Initiative