Skip to content

AIMO Standard

Agent Activity Log Format

Agent Activity Log Format

Purpose

This schema defines a vendor-neutral format for logs that document agentic AI privilege exercise, tool execution, and recursive operations. It enables organizations to:

Maintain an auditable record of autonomous agent actions
Track "who did what with what authority" for compliance and incident investigation
Support explainability for agentic AI operations in audit contexts

Event model

The schema supports four event types that capture the agentic operation lifecycle:

Event Type	Description
`agent_run`	Start or completion of an agent execution session
`tool_call`	Agent invoking a tool or external action
`tool_result`	Result returned from a tool invocation
`escalation`	Agent requesting human intervention or elevated privileges

Required fields (MUST)

Field	Type	Description	Example
`event_time`	string (ISO8601)	Timestamp of the event	`2026-01-15T09:30:00Z`
`agent_id`	string	Identifier of the agent	`agent-coding-assistant-v2`
`agent_version`	string	Version of the agent	`2.1.0`
`run_id`	string	Unique identifier for this run/session	`run-20260115-abc123`
`event_type`	string	Type of event	`agent_run`, `tool_call`, `tool_result`, `escalation`
`actor_id`	string	Initiating user or service	`user@example.com`
`tool_name`	string	Name of the tool invoked	`file_write`, `api_call`, `shell_exec`
`tool_action`	string	Action performed by the tool	`create`, `read`, `update`, `delete`, `execute`
`tool_target`	string	Target of the action	`/path/to/file`, `https://api.example.com`
`auth_context`	string	Permission/role summary	`role:developer, scope:project-x`
`input_ref`	string	Hash or URI to input (not the content itself)	`sha256:def456...`
`output_ref`	string	Hash or URI to output (not the content itself)	`sha256:ghi789...`
`decision`	string	Policy decision applied	`allow`, `block`, `needs_review`, `unknown`
`evidence_ref`	string	Reference to related evidence	`urn:evidence:...`

Optional fields (SHOULD/MAY)

Field	Type	Description
`recursion_depth`	number	Current recursion depth for nested agent calls
`retry_count`	number	Number of retries for this action
`policy_id`	string	Policy that triggered the decision
`prompt_template_id`	string	Prompt template identifier
`model`	string	Model used for this action
`latency_ms`	number	Latency in milliseconds
`cost_estimate`	number	Estimated cost of this action
`error_code`	string	Error code if the action failed

Safety notes

!!! warning "Agentic risk assumptions" When logging agentic AI activity, assume the following risks:

- **Prompt injection**: Malicious inputs may attempt to manipulate agent behavior
- **Over-privilege**: Agents may have broader permissions than intended for a specific task
- **Recursive loops**: Agents may enter unintended recursive execution patterns
- **Confused deputy**: Agents may be tricked into acting on behalf of unauthorized parties

The schema is designed to capture "who did what with what authority" to support post-incident analysis and audit explanations. It does not prevent these risks; organizations must implement appropriate guardrails.

!!! warning "Data handling" - Do not embed secrets, credentials, or sensitive content in input_ref or output_ref. - Use hash references or secure URIs to separately stored content. - Apply appropriate access controls and retention policies.

JSON Schema

Download: agent-activity.schema.json

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "required": [
    "event_time", "agent_id", "agent_version", "run_id", "event_type",
    "actor_id", "tool_name", "tool_action", "tool_target", "auth_context",
    "input_ref", "output_ref", "decision", "evidence_ref"
  ],
  "properties": {
    "event_time": { "type": "string", "format": "date-time" },
    "agent_id": { "type": "string", "minLength": 1 },
    "agent_version": { "type": "string", "minLength": 1 },
    "run_id": { "type": "string", "minLength": 1 },
    "event_type": { "type": "string", "enum": ["agent_run", "tool_call", "tool_result", "escalation"] },
    "actor_id": { "type": "string", "minLength": 1 },
    "tool_name": { "type": "string", "minLength": 1 },
    "tool_action": { "type": "string", "minLength": 1 },
    "tool_target": { "type": "string", "minLength": 1 },
    "auth_context": { "type": "string", "minLength": 1 },
    "input_ref": { "type": "string", "minLength": 1 },
    "output_ref": { "type": "string", "minLength": 1 },
    "decision": { "type": "string", "enum": ["allow", "block", "needs_review", "unknown"] },
    "evidence_ref": { "type": "string", "minLength": 1 }
  },
  "additionalProperties": true
}