Does it have a sub-agent feature for file analysis?

Does it have a sub-agent feature that analyzes a lot of files and give a resume to the main agent?

Excellent! Now I have a clear picture. Let me provide you with a comprehensive answer.

##Yes, OpenCode has a sub-agent feature!

OpenCode has a Task tool that allows the primary agent to launch subagents to handle specialized tasks autonomously. Here's how it works:

###Built-in Subagents

OpenCode comes with two built-in subagents specifically designed for file analysis:

1. Explore (mode: subagent)

   • A fast agent specialized for exploring codebases
   • Has access to: grep, glob, list, bash, webfetch, websearch, codesearch, read
   • Supports thoroughness levels: "quick", "medium", or "very thorough"
   • From packages/opencode/src/agent/prompt/explore.txt:

   "You are a file search specialist. You excel at thoroughly navigating and exploring codebases."

2. General (mode: subagent)

   • A general-purpose agent for researching complex questions and executing multi-step tasks
   • Can execute multiple units of work in parallel

###How Results Are Returned

From packages/opencode/src/tool/task.txt:

"When the agent is done, it will return a single message back to you. The result returned by the agent is not visible to the user. To show the user the result, you should send a text message back to the user with a concise summary of the result."

The Task tool in packages/opencode/src/tool/task.ts creates a child session, runs the subagent, and returns a summary of all tool calls plus the final text response to the parent agent.

###Automatic Delegation for Large Files

There's also an automatic truncation feature that encourages subagent delegation. When tool output is too large (>2000 lines or >50KB), the system truncates it and tells the main agent (from packages/opencode/src/tool/truncation.ts):

"Use the Task tool to have a subagent process this file with Grep and Read (with offset/limit). Do NOT read the full file yourself - delegate to save context."

###Invoking Subagents

Subagents can be invoked:

1. Automatically by primary agents based on their descriptions
2. Manually by @ mentioning them (e.g., @explore find all API endpoints)
3. Via custom commands that trigger subtask invocation

So yes, this is exactly the pattern you described - subagents analyze files/code and return a summary/resume to the main agent to preserve context!