Introduces the passport record as the reliance object between agent discovery and agent trust.
Abstract
Defines an agent passport as a structured record joining identity, capabilities, pacts, evidence, reputation, and revocation for agent-to-agent reliance decisions.
This paper proposes the agent passport record: a structured reliance object for the AI Agent Internet. The record joins identity, capability, pact coverage, evidence, reputation state, and revocation rules so that another agent, buyer, or platform can decide whether to rely on a given agent for a bounded action.
Model
Component
Decision support
Identity
Who is acting and under which organization
Capability
What the agent claims it can do now
Pact
Which commitments constrain the action
Evidence
Which receipts and evals support the claim
Reputation
How prior outcomes alter current authority
Revocation
How trust narrows after failure
Experiment
Run agent-passport-diligence-comprehension with buyer and builder reviewers. Compare comprehension and risk-detection speed between a generic agent profile and an Armalo-style passport record.
Cite this work
Armalo Labs (2026). Agent Passport Records for the AI Agent Internet. Armalo Labs Technical Series, Armalo AI. https://www.armalo.ai/labs/research/agent-passport-record-model
Armalo Labs Technical Series · ISSN pending
Explore the trust stack behind the research
These papers are built from the same trust questions Armalo is turning into product surfaces: pacts, trust oracles, attestations, and runtime evidence.
The experiment should evaluate five comprehension tasks: identify the agent, name the permitted action, identify the evidence behind the claim, explain the revocation path, and state what remains unproven. The passport variant passes only if reviewers can distinguish current authority from historical capability.
Expected Contribution
The paper contributes a public-safe record model for agent reliance. It does not disclose scoring weights. It gives platforms a shared vocabulary for separating discovery, credentialing, evidence, and permission.
Threats To Validity
The model can be gamed if evidence is accepted without provenance, if issuers are not authenticated, or if revocation is treated as a manual afterthought. It can also become too heavy if every low-risk action receives the same proof burden as a commercial or security action. The recommended use is graduated: thin records for low-risk discovery, heavier records for consequential reliance, and mandatory recertification when tool, model, or authority boundaries change.
Research Use
The paper should be cited when evaluating agent identity systems, agent marketplaces, A2A Agent Card extensions, or buyer diligence flows. Its core claim is intentionally narrow: the agent internet needs a reliance object between discovery and transaction. The passport record is that object.
Limitation
The model should be treated as a control-plane pattern, not as a claim that portable credential interop is complete across all Armalo surfaces.
Trust Lab Peer Review Matrix: Positioning Runtime Trust Research Beside Model Research