Paste into Claude Code / CLAUDE.md task

# DevAgentRadar → Claude Code

You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.

## Context
Assistant: Claude Code
Proposal: Set CLAUDE_CODE_TOOL_MEMORY_LIMIT for Bash tool cgroups on Linux
Summary: Opt into Linux memory cgroups for Claude Code Bash tool commands so runaway builds cannot stall the session, and wire related v2.1.233 controls (identity forwarding, WebFetch cache TTL, GitLab MR worktrees) into daily agent workflows.
Primary source: https://github.com/anthropics/claude-code/releases/tag/v2.1.233

## Why it matters
v2.1.233 adds opt-in memory cgroup support via CLAUDE_CODE_TOOL_MEMORY_LIMIT so a runaway Bash build cannot stall the session. Teams running long agent sessions on Linux should set this in their environment. The same release adds forward_user_identity on Anthropic apps-gateway upstreams for per-user spend attribution behind a proxy, CLAUDE_CODE_WEBFETCH_CACHE_TTL_MS to tune WebFetch URL cache TTL (default 15 minutes), and GitLab merge request URL support on --worktree (MRs show as !N in claude agents). Adopting these together hardens session stability, cost attribution, and MR-centric worktree flows without waiting for further releases.

## Suggested actions
1. export CLAUDE_CODE_TOOL_MEMORY_LIMIT=4G  # or your cap in bytes/suffix; add to shell profile or agent runner env so Bash tool commands run under a memory cgroup on Linux
2. In apps gateway Anthropic upstream config, set forward_user_identity: true so the signed-in user identity is sent as headers and the proxy can attribute spend per user
3. export CLAUDE_CODE_WEBFETCH_CACHE_TTL_MS=900000  # keep 15m default, or lower (e.g. 300000) for fresher fetches; set in the environment that launches claude
4. claude --worktree https://gitlab.example.com/group/project/-/merge_requests/123  # open a worktree from a GitLab MR URL; confirm it appears as !123 in `claude agents`
5. claude plugin validate  # re-run after upgrade so bare .claude/skills SKILL.md frontmatter parse failures are reported

## Config surfaces this release may change
- permission rules **(high confidence)** — check your repo before applying
- settings files — check your repo before applying
- MCP servers — check your repo before applying
- skills — check your repo before applying


## Your job
1. Restate the change in one sentence.
2. Propose a minimal plan for my repo (or a throwaway pilot).
3. Implement only what I approve; prefer small diffs and tests.
4. Call out risks (permissions, breaking APIs, cost).

Start by confirming you understood the proposal.