Anthropic

Claude Code adds a way to decline permission prompts

For unattended work, check what the job actually skipped before you trust its final report.

Source published 2026-11-06Coding automation and editorial analysisAI geeksAI-assisted editorial

Sources checked 2026-11-07. Access and availability can change.

What changed

Anthropic published Claude Code 2.1.319 on November 6. The `permissions.denyInsteadOfPrompting` setting denies a tool call that would wait for permission and lets the rest of the turn continue. [1] [2]

Check auto mode separately

Anthropic also changed auto mode for some setups with telemetry or feature flags off. If the server cannot check a tool call, Claude Code can ask or deny it instead of falling back to its built-in check. The release notes say this change leaves Bedrock, Vertex, and Foundry behavior as it was. [1]

Our take

In a spare project, give an unattended job two steps. Let it read a sample file, then give it a task that would normally ask permission before writing. Check the output for what it read, what it skipped, and what actually changed on disk. A job that reaches the end can still have left out a step your report depends on.

Sources & further reading

Open the source to check the original wording and current availability.

This is a source-linked briefing, not a hands-on product review. Provider claims and outside evaluations are identified in the text.