AI News
All stories
Official

Claude Code 2.1.325 makes scheduled tasks ask before running some skills

Scheduled tasks and /loop now ask before running a skill that declares allowed-tools or hooks, and Skill(name) must be added to permissions.allow to run it unattended.

Products · · 1 min read

Claude Code 2.1.325 changes how scheduled tasks and /loop handle skills that declare allowed-tools or hooks: they now stop and ask before running one. The changelog entry for the release says that to run such a skill unattended, a user must add Skill(name) to permissions.allow, the permission setting that pre-approves an action. Until that rule exists, a scheduled run that reaches such a skill has to wait for a person to respond.

Before this release, a scheduled task or /loop run could use such a skill with no pause. Now the user records that choice by adding the Skill(name) entry to permissions.allow.

A second change affects scheduled work too. In Desktop, IDE and SDK hosts, the "Always allow" option now offers a standalone rule only when saving that rule would allow the same command again, so a scheduled task that relied on one can now wait for the user.

Both changes mean scheduled work can pause where it used to proceed. Anyone who depends on scheduled tasks running skills without a person present will need to add the Skill(name) rule for each affected skill, or accept that those runs stop at the prompt.

What changed

Scheduled tasks and /loop now ask before running a skill that declares allowed-tools or hooks, and Skill(name) must be added to permissions.allow to run it unattended.

Type: Breaking change · Applies to: Claude Code 2.1.325, /loop, permissions.allow

Why it matters

Automated workflows that depended on silent execution of skills may stall until someone approves them, which affects reliability of scheduled AI agent jobs.

What you can do

Review scheduled tasks and /loop jobs that use skills with hooks or allowed-tools, and add explicit Skill(name) permissions only where unattended runs are intended.

Original source

Where we saw it

1 source

Claude Code changelog

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works