Rate limit error on MAX+ plan when invoking slash commands with low usage

Bug description

When invoking a custom slash command (/helpdesk), Claude Code returns:

Claude API Error: Rate limit reached

This happens despite being on a MAX+ plan with only a few percent of session usage consumed.

Steps to reproduce

1. Have a MAX+ subscription with low usage (single-digit % of monthly quota)
2. Create a custom slash command (.claude/commands/helpdesk.md) that:

• Has model: sonnet in frontmatter
• Uses allowed-tools: Bash, Read, Grep, Glob, Agent
• Auto-executes a CLI command via ! backtick syntax producing ~80 lines of output
• Has a ~130-line prompt template

1. Invoke the command: /helpdesk 000329608
2. Get Claude API Error: Rate limit reached

Expected behavior

The slash command should execute normally since the plan has sufficient quota remaining.

Actual behavior

Immediate rate limit error before the command starts processing. Retrying after a few seconds sometimes works.

Environment

• Plan: MAX+ (Claude Code)
• Model: Sonnet (specified in command frontmatter)
• Platform: Linux
• Claude Code: CLI

Notes

• The error is from Anthropic API, not from any external API called by the command
• Large CLAUDE.md context (~400 lines) may contribute to the issue
• This seems like a per-minute/per-request rate limit rather than a monthly usage limit
• The issue is intermittent — sometimes the same command works fine