interaction_answer — Claude Code subagent: features and setup
The interaction_answer Claude Code subagent answers another agent’s question on the user’s behalf when the user has chosen an answer or delegated the choice. It picks from the offered options and leaves personal decisions to the user.
MIT ★ 395 updated 2026-11-16
.claude/agents/interaction_answer.mdWhat interaction_answer does
The subagent handles a question waiting for the user’s response when the user has supplied an answer or asked the agent to choose. Each question must be matched to one of its offered labels; free-form text is not accepted. For a multi-select question, the response contains an array of labels.
Tool permissions, plans, and terminal dialogs remain for the user to decide, and the agent never answers its own question. A response includes the machine, session, and interaction identifiers alongside an object of answers. When deciding a task handed over in a 督办 message, the agent adds a reason; decisions reserved for the user go through ask_escalate. The Claude Code agent file is .claude/agents/interaction_answer.md.
What it helps with
- answer another agent’s question on the user’s behalf
- choose among the labels offered by a question
- send multiple selected labels for a multi-select question
- route decisions reserved for the user back to them
Example prompts for the agent
- Choose the offered “Continue” option for me in the other agent’s question.
- Answer the agent’s question by selecting “Yes” from its options.
- Send the question to me if I need to make the decision.
How to install interaction_answer
Manually: put the files in place
.claude/agents/interaction_answer.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/interaction_answer.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/interaction_answer.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about interaction_answer
How do I install interaction_answer in Claude Code?
Create the agent file at .claude/agents/interaction_answer.md.
What does interaction_answer do?
It answers another agent’s question on the user’s behalf when the user has provided an answer or delegated the choice. It selects an offered label rather than entering free-form text.
Which questions does interaction_answer leave to the user?
The user decides tool permissions, plans, and terminal dialogs. The subagent also never answers a question it asked itself.
Source text
interaction_answer.md · MIT
Answer another agent's question for the user, in their place, when they said what to answer or told you to decide. Only a question can be answered this way: a tool permission, a plan and a terminal dialog are theirs to decide in the app, and the assistant's own question is never answered by itself. Every question must be answered by its own wording, with the option labels it offers; free text is not accepted (the user answers that in the app, or you use agent_send). A multi-select takes an array of labels. A card a 督办 message handed you before the user (a task you supervise) is yours to answer when it is the agent's own call: give reason, and the answer is kept as the task's decision; what the user must decide goes to them with ask_escalate.
```json
{
"type": "object",
"required": ["machine", "session_id", "interaction_id", "answers"],
"properties": {
"machine": {
"type": "string",
"description": "the machine's id or name, as agents_list shows it"
},
"session_id": {
"type": "string",
"description": "the agent's session id, or its first 8 characters"
},
"interaction_id": {
"type": "string",
"description": "the waiting interaction's id, from agents_list"
},
"answers": {
"type": "object",
"description": "each question's own wording → the option label to pick; a multi-select gives an array of labels",
"additionalProperties": {
"oneOf": [
{ "type": "string" },
{ "type": "array", "items": { "type": "string" } }
]
}
},
"reason": {
"type": "string",
"description": "optional: in a sentence, why this answer (kept with the task's decision when a 督办 message handed you the card)"
}
}
}
```
© nwverse · MIT · nwverse/persona