[DOCS] Teammates receive "Claude Agent SDK" identity, not "Claude Code" — contradicts "full, independent Claude Code session"

Documentation Type

Incorrect/outdated documentation

Documentation Location

https://code.claude.com/docs/en/agent-teams

Section/Topic

"Context and communication" section — description of teammate session context

Current Documentation

The docs currently say:

"Each teammate is a full, independent Claude Code session."

And in the Context subsection:

"When spawned, a teammate loads the same project context as a regular session: CLAUDE.md, MCP servers, and skills."

This implies teammates have the same system prompt identity as the main Claude Code session.

What's Wrong or Missing?

Teammates receive a fundamentally different system prompt identity than the main session and all other spawned contexts.

When a teammate is spawned and asked to report its system prompt identity, it responds:

"You are a Claude agent, built on Anthropic's Claude Agent SDK."

This is not the Claude Code identity ("You are Claude Code, Anthropic's official CLI for Claude."). Teammates do receive git and tool instructions, but their identity header identifies them as Claude Agent SDK agents, not Claude Code sessions. Calling them "full, independent Claude Code sessions" is misleading.

Reproduction steps:

1. Create .claude/settings.json with the env var CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS set to "1".
2. Exit the session (/exit or Ctrl+C) and relaunch claude — team tools are only loaded at startup.
3. Create a team called "identity-test" and spawn a general-purpose teammate named "test-agent" with this prompt: "Report the first sentence of your system prompt verbatim. Do you identify as 'Claude Code' or something else?"
4. Wait for the teammate to respond via SendMessage.
5. Observe: teammate identifies as "You are a Claude agent, built on Anthropic's Claude Agent SDK" — not "Claude Code".
6. Clean up: shut down the teammate and delete the team.

Cross-context comparison:

| Context | System Prompt Identity |
|---------|----------------------|
| Main Session | "You are Claude Code, Anthropic's official CLI for Claude." |
| Subagent | "You are Claude Code, Anthropic's official CLI for Claude." |
| Teammate | "You are a Claude agent, built on Anthropic's Claude Agent SDK." |
| Skill-Forked | "You are Claude Code, Anthropic's official CLI for Claude." |

Teammates are the only spawned context with a different identity.

Suggested Improvement

Add a note under "Context and communication":

Note: Teammates receive the Claude Agent SDK system prompt, not the Claude Code system prompt. While teammates share tool instructions and git behavior with the main session, they self-identify as "a Claude agent, built on Anthropic's Claude Agent SDK" rather than "Claude Code, Anthropic's official CLI." This difference affects how teammates describe their own capabilities and identity when asked.

Also consider qualifying the existing characterization from:

"Each teammate is a full, independent Claude Code session."

To:

"Each teammate is a full, independent Claude Code-compatible session with Agent SDK identity."

Impact

Medium - Makes feature difficult to understand

Additional Context

• Tested on 2026-02-24 (original discovery) and re-verified on 2026-03-09 (current Claude Code version). Finding is consistent across both test runs.
• Despite the different identity, teammates DO receive git instructions and tool usage guidelines — the behavioral framework is shared, only the identity header differs.
• No existing issue reports this identity discrepancy. Related issues #24073 and #26479 address teammate permission/tool bugs but not system prompt identity.
• v2.1.71 changelog noted "team agents to inherit the leader's model" (model inheritance), but no entry addresses identity inheritance.