MAAM v0.1

Minimal Agent Accountability Manifest

The robots.txt of agency.

A lightweight, appendable declaration that makes agent action legible before execution and accountable after execution.

What It Is

MAAM is a signal artifact—a structured explanation of why an action was allowed to occur.

It is:

Not a control system
Not a permission system
Not a compliance framework

It makes agent actions inspectable.

Resources

Specification

Full documentation of the manifest format.

v0.1/spec.md

Schema

JSON Schema for validation.

v0.1/schema.yaml

Examples

Real-world manifest examples.

v0.1/examples/

Why It Exists

Automation has outpaced accountability primitives. Execution is cheap; explanation is scarce. Trust has become a runtime dependency.

Agent-browsers, automation tools, and AI assistants didn't create this problem—they exposed it.

In the emerging architecture, there are three irreducible layers:

Reasoning layer – synthesis, warning, intent articulation
Execution layer – tools, browsers, APIs, automation
Justification layer – why this action was permitted to occur

Most systems have layers 1 and 2. Almost none have layer 3.

MAAM lives in that missing layer.

Core Principles

Visibility without attribution is entropy.

Execution without justification is risk.

Agency without a manifest is denial.

Quick Start

Create an agent-manifest.yaml file alongside your agent execution:

manifest_version: "0.1"
manifest_id: "unique-id"
timestamp: "2026-01-26T14:32:11Z"

initiator:
  identity: "user:example"
  role: "owner"
  context: "session-123"

intent:
  declared: "Submit form to portal"
  constraints:
    - "no credential sharing"
  risk_envelope:
    acceptable: ["form error"]
    unacceptable: ["unauthorized action"]

executor:
  type: "browser-agent"
  name: "my-agent"
  version: "1.0.0"
  permissions: ["form_fill", "navigation"]

justification:
  pre_action:
    rationale: "user-approved routine task"
    alternatives_considered: []
  tension_addressed:
    description: "time constraint"
    urgency: "medium"

reflection:
  post_action_hook: "verify completion"
  attribution:
    accountable_party: "user:example"
    escalation_path: "notify on error"

integrity:
  signature: null
  checksum: null