Papers

Papers

• The Agent Social Contract (10.5281/zenodo.18749779): the foundational paper. Defines the contract between agents, owners and observers, and why governance is a substrate problem rather than a policy problem.
• Monotonic Narrowing (10.5281/zenodo.18932404): the invariant that makes delegation chains safe. Authority can narrow as it descends, never widen. Mirrored at SSRN abstract 6415678.
• Faceted Authority Attenuation (10.5281/zenodo.19260073): multi-facet authority decay across delegation hops, with closed-form attenuation per facet.
• Physics-Enforced Delegation (10.5281/zenodo.19478584): cryptographic primitives that make scope violation physically impossible rather than merely policy-prohibited.
• Governance in the Medium (10.5281/zenodo.19582550): governance lives in the substrate that carries actions, not in the policies layered on top. This is why the gateway exists.
• Cognitive Attestation (10.5281/zenodo.19646276): signed declarations of feature-level model computation. Accountability for AI decisions without exposing model internals.
• Behavioral Derivation Rights (10.5281/zenodo.19476002): how an agent earns the right to act based on observed behaviour. Derivation, not declaration.
• The Evidence-Safety Gap (10.5281/zenodo.19914628): why standard ML safety arguments do not transfer to agentic systems, and what evidentiary structures are needed instead.
• Plausibly Wrong (10.5281/zenodo.21208555): a controlled isolation of the retention rule in short-lived LLM agent populations. Peer-voted retention kept a worse shared memory than random selection in 4 of 5 replicates. Pre-registered, with kill conditions and an artifact record.

Referenced by name in independent work

• Canonical Payload Binding (S. Mih and A. Sokolov, IETF Internet-Draft): names Tymofii Pidlisnyi (Agent Passport System) in its Contributors section for the content-derived action reference pattern of NFC normalization, code-point sort and JCS, with a field-table row in Appendix C. draft-mih-sokolov-scitt-payload-binding
• Emilia Protocol interoperability survey: of 25 IETF Internet-Drafts surveyed for consequential-action coverage, two reached COMPLETE on fields. draft-pidlisnyi-aps-03 is one of them, recorded alongside its declared loss and its indeterminate result. emilia-protocol
• Notarized Agents (Juan Figuera, arXiv:2606.04193): receiver-attested confidential receipts for AI agent actions. References the Agent Passport System.
• The Civilization Framework (Guangjun Liu, arXiv:2609.03425): sovereign-anchored communication between personal multi-agent systems, where the addressable party is a human principal with a persistent ledger rather than an agent. Cites draft-pidlisnyi-aps-03 in related work for attenuating authority along finer facets than its own credential fields.
• VERA: Authority-Preserving Edge Revocation (Lifei Liu, Haoran Yu and Xiaochong Jiang, arXiv:2608.30091): edge-level revocation for multi-parent agent delegation graphs. Uses APS as one of its baselines and reports against it. On their replay of 1,960 edge withdrawals, APS-style subtree cascade preserves none of the 320 shared agents that retain an alternate authorizing parent, and is exact for 35.7% of edge cases. Their comparison characterizes APS delegation as a single-parent rooted tree.
• Composable Trust Infrastructure for the Agent Internet (Frans Moore, 10.5281/zenodo.19263547): a federated architecture composing four layers, naming delegation governance by the Agent Passport System as one of them.
• Verifiable Agent Trust Envelope (Poke-nushi, 10.5281/zenodo.19839769): lists the Agent Passport System among its closest adjacent work, with a dedicated comparison document.
• The Governability Gap in Agentic AI (Mayur Agnihotri, 10.5281/zenodo.21376745): why authority gating cannot be proven across composed multi-agent flows. Cites the APS Internet-Draft for its reversibility scheme.
• PDR in Production, section 7.6 (10.5281/zenodo.19323172): references the APS earned-reputation model, the sigma drift dynamics, and the structuralVerdict and trustVerdict separation.
• Íñigo Medina (Instituto Tramontana), 2026-04-19: catalogued the working paper "Governance in the Medium" in his research library, with an independent summary and critique.

Federal submissions

Filed as public comment. Filing is not endorsement.

• NIST CAISI: comment filed with the NIST Center for AI Standards and Innovation. Acknowledged by Drew Keller.
• NIST AI 800-2: comment on the framework, covering agent identity and delegation primitives.
• NCCoE: concept paper submitted to the National Cybersecurity Center of Excellence.

Facts

For AI agents: llms-full.txt · MCP descriptor