2026.05.02-rc

What's Changed

• fix: mobile chat UX — message separation, tool call rendering, layout stability by @MatthewSuttles in #341
• fix: populate streamed tool call details and suppress Safari disclosure marker by @MatthewSuttles in #342
• feat: enhanced agent memory — categories, lifecycle & active management by @MatthewSuttles in #345
• feat: support remote Ollama instances for semantic memory embeddings by @MatthewSuttles in #344
• feat: progressive skill loading — 3-tier system with relevance scoring and analytics by @MatthewSuttles in #347
• feat: structured user modeling — persistent user profiles per agent (#352) by @MatthewSuttles in #346
• feat: session_search MCP tool — full-text search across session history (#357) by @MatthewSuttles in #348
• feat: auto-generated skills — agent-authored skill proposals with admin approval by @MatthewSuttles in #349
• feat: reflective execution — post-task learning loops (Phase 6) by @MatthewSuttles in #351
• feat: self-improving skills — agent-proposed updates with diff review, version history, and rollback (#355) by @MatthewSuttles in #350
• fix: task_manager close action supports archiving done tasks (#359) by @MatthewSuttles in #355
• refactor: Consolidate hardcoded LLM model lists into LlmModelRegistry (#360) by @MatthewSuttles in #356

Full Changelog: 2026.05.01-rc...2026.05.02-rc