ToolCallContext
The evaluation input provided to policies at the Tool Call Interception point.
Schema ID: https://agentpolicyspecification.github.io/schemas/tool-call-context.schema.json
Properties
| Property | Type | Required | Description |
|---|---|---|---|
tool_name | string | Yes | The name of the tool the LLM has requested to invoke |
arguments | object | Yes | The arguments provided by the LLM |
calling_message | AssistantMessage | Yes | The assistant message that produced the tool call |
metadata | Metadata | Yes | Agent and session context |
AssistantMessage
| Property | Type | Required | Description |
|---|---|---|---|
role | "assistant" | Yes | Always "assistant" |
content | string | Yes | The text content of the message |
Metadata
| Property | Type | Required | Description |
|---|---|---|---|
agent_id | string | Yes | Unique identifier for the agent |
session_id | string | Yes | Unique identifier for the session |
timestamp | string (date-time) | Yes | ISO 8601 timestamp of the interception |
Example
json
{
"tool_name": "read_file",
"arguments": { "path": "/workspace/data.csv" },
"calling_message": {
"role": "assistant",
"content": "I will read the file to answer your question."
},
"metadata": {
"agent_id": "agent-1",
"session_id": "session-abc",
"timestamp": "2026-03-31T10:00:01Z"
}
}