v0.4.0

What's Changed

• docs(changelog): pin the 0.3.0 compare link to the v0.3.0 tag by @Folken2 in #58
• feat: Agent Plugin Registry (Agent Plugins v1.0.0 support) by @markfolken in #65
• feat: nuvel mcp serve — MCP server exposing Nuvel Skills as resources and tools by @markfolken in #67
• feat: add bot fleet management by @markfolken in #69
• feat: evalv2 — skill-driven AI-native evaluation (all 4 phases) by @markfolken in #66
• feat: nuvel mcp serve --theme — role-scoped skill discovery by @markfolken in #68
• fix: correct migration header and ship THIRD_PARTY.md in wheel by @markfolken in #70
• fix: add halt acknowledge tool and wire command_safety guard by @markfolken in #71
• fix: replace plugin parser with AST walker, add subsystem coverage guard by @markfolken in #72
• docs: add bot fleet management section and 0.4.0 changelog by @markfolken in #73
• feat: add Buzz backend scaffold (--framework buzz) by @markfolken in #74
• feat: rename nuvel new → nuvel agent create, drop Claude backends by @markfolken in #75
• chore: bump google-adk lock to 2.10.0 + validate breaking-change surface by @markfolken in #84
• docs: keep LazySkillToolset — ADK 2.10 native change-detection does not supersede it by @markfolken in #85
• fix: forward process_llm_request in LazySkillToolset (skill guidance was silently dropped) + Wave 3 disclosure evaluation by @markfolken in #86
• fix: make nuvel's plugins, guardrails and wheel safe for a shared server by @Folken2 in #87
• docs: VISION.md — adopt the commodity, build the differentiator (layer ownership + roadmap refocus) by @markfolken in #88
• feat(mcp): hub search, browse, install and durable feedback in the Skills MCP server by @markfolken in #92
• feat(mcp): reconcile the skill feedback loop with the hub server — one feedback store by @markfolken in #93
• chore(release): 0.4.0 — unblock PyPI publish (mcp, bots, fleet) by @markfolken in #95

Full Changelog: v0.3.0...v0.4.0