Claude Code: Auto-Mode Now Shows Why Commands Were Denied
Claude CodeWhen Claude Code runs in auto mode and the permissions classifier blocks a command, the interface now surfaces the reason it was denied. The explanation appears in the transcript, the denial toast, and the recent denials list under /permissions. The change shipped in v2.1.193.
Sources & Mentions
2 external resources covering this update
Auto mode now explains its denials
Claude Code's auto mode runs commands without stopping for per-action approval, relying on a permissions classifier to block anything unsafe. Version 2.1.193 makes those blocks transparent: when a command is denied, Claude Code now records the reason it was denied.
What it does
The denial reason now appears in three places: the session transcript, the denial toast that pops up at the moment of the block, and the recent denials list under /permissions. Instead of a bare "denied" message, you see why the classifier rejected the command, for example a destructive git operation you did not ask for or a command matched by a deny rule.
Why it matters
Silent or unexplained denials are confusing. You are left guessing whether a rule, a safety heuristic, or a configuration issue stopped the command, and that slows down debugging your permissions setup. Surfacing the reason turns each denial into actionable feedback. You can decide whether to adjust a rule, rephrase the request, or leave the protection in place.
Who benefits
Teams and individuals who run Claude Code in auto mode for speed but still want to understand and tune what gets blocked. It is particularly helpful for administrators refining permission policies, since the reasons appear directly in /permissions alongside the denials they need to review.