Claude Code Loop Skill

A built-in Claude Code skill that schedules recurring prompts inside a live session using cron-like intervals or one-shot reminders. Implemented via cron_create, cron_list, and cron_delete tools. Works in terminal, VS Code, and desktop app. It is best for "help me today" tasks because it depends on an active local session, has expiry and jitter constraints, and has no durable catchup.

Key points

• Invoked via /loop or natural language ('every 5 minutes, check X')
• Tools: cron_create, cron_list, cron_delete
• Expiry is product-surface dependent in Nate's May 2026 testing: desktop showed a 3-day horizon, while terminal loops exposed a 7-day horizon ^[src-086]
• Dies when the terminal or session closes — no persistence
• Terminal loops can survive /clear, and users can even schedule a recurring /clear loop to reduce context rot during long work ^[src-086]
• Recurring runs may have jitter of up to about 30 minutes, so /loop is not a precise scheduler ^[src-086]
• Runs in the same session, so loops share context (pro and con)
• Available in terminal, VS Code, and desktop app
• Use case: short-lived polling (email, deploys, logs) vs long-term automations
• Boris Cherny says loops are becoming central to his own workflow: he uses loops to babysit PRs, fix CI, auto-rebase, keep CI healthy, fix flaky tests, and cluster feedback every 30 minutes ^[src-054]
• Boris describes /loop as "the simplest thing that works" and says loops feel like the future, especially as models start choosing recurring loops naturally when they notice changing data ^[src-054]
• Anthropic's scientific-computing article positions /loop beside Ralph and GSD-style scaffolds as ways to keep long-running agents engaged until a measurable success criterion is actually met ^[src-072].
• The pattern is especially useful when an agent prematurely claims completion on complex multi-part work despite failing tests or accuracy targets ^[src-072].

Related entities

• Claude Code
• Boris Cherny

Related concepts

• Claude Code Scheduled Tasks
• Claude Code Cloud Routines
• AI-Native Organizational Process
• Agent Teams
• Ralph Loop Orchestration
• Long-Running Scientific Agents
• Agent Deployment Modes

Source references

• ^[src-004] Nate Herk cluster — Nate Herk — Claude Code cluster (21 videos)

– Videos referenced: OUyfxhFtGCo

• ^[src-054] Sequoia Capital — "Anthropic's Boris Cherny: Why Coding Is Solved, and What Comes Next" (2026-05-04)
• ^[src-072] Siddharth Mishra-Sharma – "Long-running Claude for scientific computing" (2026-03-23)
• ^[src-086] Nate Herk — "I Tested 3 Ways to Deploy Claude Agents (Here's When to Use Each)" (2026-05-15)