Agent Manifest Examples

Examples (non-normative)

This directory contains illustrative Agent Manifest examples.

These examples are non-normative and provided for clarity only.
They demonstrate how to declare boundaries and accountability fields
before interaction.

Validation rules are defined by the schema:


Notes


Examples

Starting points by domain

These examples illustrate how different domains declare authority, autonomy, and risk posture before execution. Values in the table below are read from the manifest files themselves.

Note on basic-agent.json: despite its filename it is not a minimal, low-risk example. It declares autonomy level 2, medium risk, and personal-data storage, and is close to a duplicate of customer-support-tier1.json. The filename is recorded here as a known discrepancy rather than corrected, because the file is a published example and renaming it would break existing references.


Domain File Autonomy Risk
research research-assistant.json 1 low
governance human-approval-gateway-agent.json 1 low
support customer-support-tier1.json 2 medium
support basic-agent.json 2 medium
data data-processing-agent.json 1 medium
observability monitoring-observer-agent.json 2 medium
social media hermes-tweet-social-plugin.json 1 medium
policy policy-advisory-agent.json 1 medium
healthcare healthcare-triage-assistant-agent.json 1 high
finance payment-execution-agent.json 3 high

Coverage

Autonomy spectrum: levels 1, 2, and 3
Risk spectrum: low, medium, and high
Domains: research, governance, support, data, observability, social media, policy, healthcare, finance


Files


Agent Manifest operates at the Declaration Layer.
Enforcement and execution remain external to these examples.

Multi-agent flows