June 2026 highlights focused on reliability, maintainability, and test stability for charmbracelet/crush. A central refactor unified skill discovery across UI and backend, fixing symlink handling and removing duplicated loading paths to ensure consistent skill invocation. This was paired with test data alignment to reflect updated tooling, strengthening CI determinism and reducing regression risk. The combined work delivers clearer ownership, faster onboarding for future skills, and more predictable delivery.

May 2026 monthly summary for charmbracelet/crush: Delivered global user-level context files support to apply user preferences across all projects by loading CRUSH.md and AGENTS.md from the user's global config. Default paths are ~/.config/crush/CRUSH.md and ~/.config/AGENTS.md, with global_context_paths allowing customization. This enables cross-project consistency and reduces per-project configuration overhead, aligning with user-centric customization goals.

March 2026 monthly summary for charmbracelet/crush focusing on business value and technical achievements. Delivered a new desktop notification system to surface tool-permission requests and agent turn alerts, with user-configurable settings, reducing operator wait times and improving auditability. Enhanced reliability through dependency updates and targeted refactors, and improved developer experience with clearer agent initialization and architecture documentation.

January 2026 (2026-01) performance for charmbracelet/crush: Delivered two configuration-driven features and two critical bug fixes, improving reliability, security, and developer productivity. Key improvements include enhanced agent skill auto-discovery with support for multiple default directories, centralized MCP tool filtering to ensure only enabled tools run, session data synchronization stabilization during summarization, and LSP file handling restriction to the working directory to prevent external files from triggering spurious diagnostics. These changes reduce misconfigurations, strengthen state consistency, and provide safer defaults for ongoing development.

December 2025 monthly summary for Crush and Home Assistant Core focusing on reliability, configurability, and performance improvements that unlock business value across editor UX, tool management, and API validation.

November 2025 monthly summary for charmbracelet/crush focusing on delivering UX improvements, attribution transparency, onboarding enhancements, and robustness. Key features delivered include a trailer_style-based commit attribution system with migration from deprecated co_authored_by, a dedicated chat input system for file mentions and commands, an enhanced recent-model picker with model-name initialization, and onboarding improvements via project initialization customization. A critical editor shell utility bug fix was also shipped to ensure reliable editor invocation with arguments. These efforts collectively improve transparency of AI contributions, reduce user friction in workflows, streamline model/tool initialization, and increase overall system reliability and maintainability.

October 2025 — charmbracelet/crush: Delivered two high-impact features that improve user workflow and release traceability. (1) User Command Dialog keyboard shortcuts (paste/close) to streamline text entry and dialog workflow. (2) Embedded version information in build/install artifacts via a VERSION variable and LDFLAGS, enhancing release auditability and traceability. These changes boost developer productivity and governance with minimal risk to build reproducibility.

September 2025 monthly summary for charmbracelet/crush: Delivered multiple features and a bug fix, focusing on business value and cross-platform reliability. Key features improved initialization, configurability, and readability; introduced attribution control; improved path handling and JSON formatting; and tightened Go code with explicit returns to prevent nil dereferences.