Copilot CLI v1.0.106-9 is a GitHub release (pre-release) published 2026-11-04 with 15 listed changes. Summary and full notes below, extracted from the official notes.
One free email: every breaking change, retirement and price change across 23 coding agents. No spam, unsubscribe anytime.
Subscribe free← All Copilot CLI releases and breaking changes
**Fixed** - apply_patch no longer counts one extra line for each file it adds or deletes, so the timeline summary, permission prompt, /usage and exit summary show the real line counts - `copilot config` now warns when a value of the wrong type makes settings.json ignored, instead of reading every setting as unset and reporting writes into that file as if they took effect - Deleting text in front of Chinese, Japanese or Korean characters in the prompt no longer leaves the line garbled until the next edit. - Markdown output no longer turns colon-delimited text such as file.ts:100:5 or crate::id::Id into emoji. - A top-level SKILL.md in a skills directory now produces a warning in /skills explaining that subdirectories are resources, not separate skills, and how to load them. - A message sent while the agent is working that mentions a file or directory (`@path`) is no longer rejected with "Submission failed" when an earlier one is still waiting. - /keep-alive busy releases sleep prevention once all tracked sessions finish working, including background turns and standalone local-shell commands - Git URL caches use short paths and diagnose path-length failures. Old caches are copied, not moved, and kept for older CLIs; delete the originals manually when no longer needed. - ACP session/prompt and session/cancel stop only the active turn, keeping background tasks and shell commands running - Long conversations now recover from oversized requests by trimming historical reasoning and retrying. - Prompt before out-of-workspace access through dangling symlinks - Rewinding and resubmitting a prompt now keeps image and file attachments. - Sessions recover from unavailable program endpoints and continue answering prompts - Skill calls explain when model invocation is disabled instead of reporting the skill as not found - /memory now reports when Memory is disabled for the session by an experiment and clarifies /memory on behavior.