CLAUDE CODE V2.1.71 ADDS /LOOP AND CRON-STYLE SCHEDULING FOR HANDS-FREE AGENT RUNS
Anthropic shipped Claude Code v2.1.71 with /loop and cron-like scheduling for recurring agent tasks, plus a wide set of stability fixes. The release adds a new...
Anthropic shipped Claude Code v2.1.71 with /loop and cron-like scheduling for recurring agent tasks, plus a wide set of stability fixes.
The release adds a new /loop command to run prompts or slash commands on an interval (for example: /loop 5m check the deploy) and cron scheduling tools for recurring prompts within a session. It also makes voice push-to-talk rebindable and expands the Bash auto-approval allowlist with common UNIX tools. See the full notes on the GitHub releases page: anthropics/claude-code v2.1.71.
Stability work targets long-running usage: fixes for stdin freezes, voice-mode startup stalls, OAuth token-refresh storms, plugin persistence across instances, plan-file collisions when forking sessions, and image handling in the Read tool. If you rely on agents that run in the background, this update aims to keep them alive and predictable.
If you want prebuilt routines to drop into loops, the community repo sickn33/antigravity-awesome-skills hosts 1,000+ agent skills you can adapt for deploy checks, debugging, and ops hygiene.
You can now schedule Claude Code to run checks on a timer or cron, turning ad‑hoc prompts into reliable background jobs.
The stability fixes reduce flakiness in long sessions, which is critical for agents that watch deploys, pipelines, or logs.
-
terminal
Create a /loop (e.g., every 5 minutes) to verify deploy health or recent error rates; observe stability across sleep/wake and network changes.
-
terminal
Use cron scheduling to align with business windows; add guardrails (max turns, timeouts) and measure cost, rate limits, and cancellation behavior.
Legacy codebase integration strategies...
- 01.
Map /loop jobs to existing cron/CI/Airflow checks; decide which tasks shift to Claude vs. stay in current schedulers.
- 02.
Review the expanded Bash auto-approval allowlist against security policy and harden where needed.
Fresh architecture paradigms...
- 01.
Stand up lightweight agent sentries for deploys and ETL freshness without new infra; persist outputs to your existing logs or tickets.
- 02.
Bootstrap routines from the antigravity skills repo to standardize prompts for debugging, testing, and triage.