DesireCore v10.0.26
Release Date: 2026-04-09

New Features
- Built-in Global Skill Expansion: Added new built-in global skills including Skill Creator, Word, PDF, Excel, PowerPoint, Web Access, and Front-End Design. Agents can now more naturally handle office documents, online search, page design, and skill creation tasks.
- Explore Delegation Preset: Agents can now delegate a read-only "Explore" workflow with one click, making it better suited for research tasks such as file search, code reading, and content analysis.
- Runtime Environment Management: A new Runtime Environment tab has been added to the Resource panel, where you can view runtime details and available tools for easier understanding and management of the agent's execution environment.
- Dedicated SendUserMessage Cards: Messages proactively sent to you by agents are now displayed as separate cards, with support for Markdown, attachments, and proactive notification indicators for a more intuitive reading experience.
- All-New Brand Visuals: The app has been updated with a new brand identity and iconography, creating a more consistent overall interface experience.
Fixes
- Fixed inaccurate Token usage statistics under some models. Usage shown in activity logs, conversation state, and single-query summaries is now more accurate and reliable.
- Fixed overcounting caused by legacy cumulative Token data in historical conversations. After upgrading, old data will also be corrected and displayed in a more reasonable way.
- Fixed an issue where usage data in activity logs was lost when clearing a conversation. Audit usage records are now preserved even after chat content is cleared.
- Fixed an issue where cleared conversations could be reused by the same agent. New conversations will no longer mistakenly use those historical sessions.
- Fixed an issue where SendUserMessage cards would only appear after the full message had finished generating. They now appear immediately during generation and continue updating in real time.
- Fixed an issue where some built-in skills were difficult to trigger in Chinese-language scenarios. Requests related to office documents and similar tasks are now more easily recognized and invoked correctly.
Improvements
- Optimized Token display in query summaries and the runtime status panel. Input, output, cache reads and writes, and cost information are now distinguished more clearly.
- Upgraded the Agent model configuration structure for compatibility. Importing and reading legacy agent configurations is now smoother and less prone to errors caused by changes in configuration fields.