Claude Code 2.1.132 team conventions

The situation

Claude Code 2.1.132 is a small release, but it matters for teams that share repo context, depend on terminal stability, and review agent work. The changelog points to a clear pattern in Claude Code team conventions: better session state, cleaner interruption handling, and less fragile terminal behavior.

That matters most when Claude Code is part of a team workflow, not a solo setup. If your group uses CLAUDE.md for durable repo memory, hooks for deterministic checks, and MCP for external tools, then session boundaries and terminal recovery affect whether work is easy to trust and resume.

The release also adds a new environment variable for Bash subprocesses and an opt-out for the alternate-screen renderer. Those controls help teams standardize how Claude Code behaves across terminals, IDEs, and remote sessions.

If you are running a Claude Code workshop or rolling out Claude Code for teams, this release is mostly about reducing friction where conventions break: paste handling, resume behavior, terminal rendering, and interrupted sessions. For more on the team setup, see Claude Code team conventions.

What changed and what to try

1. Start with the two changes that affect shared-repo work most directly: CLAUDE_CODE_SESSION_ID is now passed into Bash subprocesses, and CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN=1 keeps the conversation in native scrollback instead of the fullscreen alternate-screen renderer. The first helps hooks and subprocesses tie work to the parent session; the second gives teams a simpler terminal mode when they want easier auditability and copy/paste behavior.

2. Tighten repo memory. If your team uses CLAUDE.md, keep it short and explicit: durable conventions at the root, narrower rules in nested files, and task-specific guidance elsewhere. Claude Code’s memory model is built around persistent instructions plus auto memory, so the useful question is not how much to add, but what every session should inherit without re-explaining.

3. Use hooks for deterministic boundaries, not vague policy. The session and subprocess changes make hooks more useful because they can anchor work to a stable session identifier. A simple CLAUDE.md fragment can look like this:

# CLAUDE.md

- Follow repository conventions before proposing new patterns.
- Prefer small diffs and reviewable steps.
- If a task touches generated files, explain why in the PR summary.
- Use hooks for validation and logging; do not rely on chat text for enforcement.

4. Keep skills focused. Claude Code docs treat CLAUDE.md as persistent instruction memory and skills as on-demand capabilities. A useful Claude Code training or workshop exercise is to write one skill for a repeated team workflow, then check whether it belongs in memory, a skill, or a hook.

---
name: repo-review-checklist
description: Check Claude-authored diffs for repo conventions, test coverage, and risky edits before review.
---

# Repo review checklist

- Confirm the diff matches the repository’s local conventions.
- Check whether the change needs a hook, permission update, or memory update.
- Verify tests or validation steps are included.

5. Revisit review habits after this release. The changelog fixes failure modes that can make agent work look worse than it is: abrupt exits on SIGINT, resume failures after truncated tool errors, blank screens after sleep or Ctrl+Z, and paste bugs that can swallow commands or inject stray escape sequences. Reviewers should now ask a simpler question: did the session end cleanly, and can it be resumed without losing context?

6. If your team uses MCP, keep permission review separate from the model prompt. Claude Code’s MCP docs frame connectors as a scope and trust boundary, which means the right review artifact is a checklist, not a paragraph of reassurance. The same applies to hooks: they are infrastructure, so document them like infrastructure.

7. Methodology note: in the Review step, check the session boundary and the repo artifact together. If a Claude-authored change depends on a hook, a memory update, or a connector, verify the file change and the operational consequence in the same pass.

Tradeoffs and limits

This release does not change the core governance problem: teams still need to decide what belongs in CLAUDE.md, what belongs in a skill, and what should be enforced by hooks or MCP permissions. The new session ID helps with traceability, but it is not a substitute for clear repo conventions.

The alternate-screen opt-out is useful, but it is a preference, not a universal default. Some teams will want fullscreen rendering for focus; others will prefer native scrollback for auditability and easier terminal recovery. Standardize the choice per team, not per person.

The fixes around paste handling, cursor movement, and terminal rendering reduce operational noise. They do not remove the need for validation, code review, or explicit permission boundaries.

If your team already has a large CLAUDE.md, this release is not a reason to add more. The better move is usually to trim, scope, and separate concerns so the memory file stays readable and the hooks stay deterministic.

Further reading

• https://code.claude.com/docs/en/changelog
• https://code.claude.com/docs/en/overview
• https://code.claude.com/docs/en/memory
• https://code.claude.com/docs/en/hooks
• https://code.claude.com/docs/en/mcp
• https://code.claude.com/docs/llms.txt
• /topics/team-conventions