Monitors Claude Code usage and stops before the window threshold is reached.
npx skills add https://github.com/nikhilsitaram/claude-caliper --skill usage-guard
Please help me install this Agent Skill. Skill: usage-guard Source: https://github.com/nikhilsitaram/claude-caliper/tree/main/skills/usage-guard Command: npx skills add https://github.com/nikhilsitaram/claude-caliper --skill usage-guard
This Skill prevents Claude Code sessions from getting rate-limited mid-task by watching usage-window consumption and stopping cleanly before the threshold.
Use the usage-guard skill to keep working on my current task, stop at 99 percent of the current usage window, and queue the remaining work if needed.
High-intent search queries and answers about installing and using this skill.
To prevent Claude Code rate limits during autonomous work, use a monitoring skill that tracks consumption and stops cleanly before the threshold. It breaks long-running tasks into checkpoints and rechecks usage after each chunk.
Yes, you can queue and resume Claude Code tasks across multiple usage windows. The skill stops at the threshold, packages remaining work into a refreshed queue state file, and records interruption breadcrumbs for safe continuation into the next window.
Yes, you need jq installed to monitor Claude Code usage limits with this skill. jq is a required dependency used by the scripts to parse and update the queue state file and checkpoint data.
The best way to guard both 5-hour and 7-day Claude Code usage caps is to apply threshold branching control flow. The same monitoring loop protects the default rolling block and the weekly cap by stopping before the threshold.
Checkpoint-based usage monitoring works by breaking a continuous work loop into chunks, updating progress, and rechecking usage-window consumption after each chunk. It applies threshold branching to stop cleanly before hitting the rate limit.
virtengineCoordinate wave-based task execution across subagents with STATE.md tracking.
cshyangResume paused GSD work sessions from a saved state.
WillHsiaoNYCWrap long-running AI coding tasks in rate-limit guardrails with checkpoints.
latestaiagentsImplement token bucket, sliding window, and concurrency rate limiting for LLM APIs.
