Release v4.0.5

[4.0.5] - 2026-03-29

Added

• make CodeClaw platform-agnostic — remove CLAUDE.md coupling
• add mandatory Agent Teams sections to all 9 skills
• enforce CLAUDE.md as authoritative source of truth across all skills

Fixed

• remove hardcoded isolation worktree from Agent Teams role tables

v4.0.4 — Platform-Agnostic Refactoring

What's Changed

Added

• Make CodeClaw platform-agnostic: remove CLAUDE.md coupling from all 9 skills
• Add load_config() to common.py merging project-config.json with CLAUDE.md fallback
• Add skills.sh as canonical platform-agnostic installer
• Extract Agent Teams documentation to docs/agent-teams.md
• Add development_branch, staging_branch, production_branch to project-config.example.json

Changed

• Replace ${CLAUDE_PLUGIN_ROOT} with ${CLAW_ROOT} across all files
• Replace "CLAUDE.md IS LAW" with "Project configuration is authoritative" in all skills
• Remove mandatory Agent Teams sections from all 9 skills
• Default install platform changed from claude-code to generic
• Update skill_helper.py, release_manager.py, test_manager.py to use load_config()
• Mark CLAUDE.md template as optional
• Remove obsolete worktree code and tests
• Trim docs and coverage snapshots

Full Changelog: v4.0.3...v4.0.4

v4.0.3

What's New

Added

• Comprehensive local code quality gate with auto stack detection (QLTY-0001)

Changed

• Refactor project structure and enhance agent memory management
• Consolidate shared utilities into scripts/common.py
• Condense skill shorthand tables and yolo definition into CLAUDE.md

Full Changelog: v4.0.2...v4.0.3

v4.0.2 - Bug fixes & Worktree Improvements

[4.0.2] - 2026-03-20

Added

• Enable worktrees by default with shared memory integration (VMEM-0052)
• Auto-create GPU env vars and retry before fallback to CPU (VMEM-0001)
• Log vector search query results to file (VMEM-0051)

Fixed

• Cross-platform file locking for config writes (RPAT-0001)
• Download timeout and integrity checks for model fetching (RPAT-0004)
• GPU path allowlist for config-loaded library paths (RPAT-0002)
• Validate and auto-download embedding model on config change (VMEM-0050)
• Single release branch when worktrees disabled (BRNCH-0001)
• Silent GPU-to-CPU fallback detection and auto-configure (FIX-0047)
• Worktree-aware root resolution for memory subsystem (VMEM-0002)
• Singleton release state issue enforcement (FIX-0048)
• Agent worktree isolation when project worktrees disabled (FIX-0049)

Security

• Search log security controls and privacy documentation (RPAT-0003)

Release v4.0.1

[4.0.1] - 2026-03-19

Added

• Auto-detect GPU hardware for ONNX Runtime setup (VMEM-0046) (#171)
• Enable claw@dnviti-plugins in settings.json

Fixed

• Worktree-aware vector memory path resolution and [BETA] opt-in (FIX-0001) (#172)

Release v4.0.0

What's Changed

• ci: make dependency review non-blocking (requires repo-level dependency graph)
• ci: fix CI failures — add requirements.txt, update action SHAs
• ci: configure CI pipeline for Python 3.12 with lint, syntax check, and pytest
• chore(release): v4.0.0
• chore: add project config, workflows, icons, and utility scripts
• chore: bump plugin version to 4.0.0
• fix: resolve RLM model recommendations collision and restore export_context
• fix: resolve vector memory and MCP review findings
• fix: resolve memory subsystem review findings
• fix: resolve skill definition review findings
• fix: resolve config, utility, and provider review findings
• fix: resolve RLM and ollama review findings
• fix: resolve task/docs/test manager review findings
• fix: add missing timeout parameter to FileLockBackend.init
• chore: bump plugin version to 4.0.0
• fix: resolve RLM model recommendations collision and restore export_context
• fix: resolve vector memory and MCP review findings
• fix: resolve memory subsystem review findings
• fix: resolve skill definition review findings
• fix: resolve config, utility, and provider review findings
• fix: resolve RLM and ollama review findings
• fix: resolve task/docs/test manager review findings

Full Changelog: v4.0.0-staging...v4.0.0

v3.5.2

[3.5.2] - 2026-03-18

Added

• opt-in vector memory MCP with seamless setup installation (VMEM-0029) (#105)

Fixed

• migrate MCP server to FastMCP API and harden .gitignore entries
• enable Ollama tool call offloading with 0-10 level control (OLLAM-0030)

Full Changelog: v3.5.1...v3.5.2

v3.5.1

What's New in v3.5.1

Added

• Platform release state sync (REL-0028) — In platform-only mode, release-state.json is stored as a ctdf-release-state GitHub/GitLab issue, shared by all collaborators without needing git pull
• Ollama tool calling (OLLAM-0027) — Full /api/chat loop with tool call support; Ollama can now invoke tools and feed results back in a multi-round conversation
• Mandatory vector memory (VMEM-0024) — Always-on semantic indexing via LanceDB; auto-indexed on every file edit via PostToolUse hook; MCP server for agent retrieval
• Configurable Ollama offloading level 0–10 (OLLAM-0023) — Granular control over which tool calls are routed to local models
• Setup wizard: memory + social config (SETUP-0026) — Interactive configuration for vector memory and social media announcement platforms

Fixed

• NFKC Unicode normalization (RPAT-v3.5.1-3) — Exclude patterns are now NFKC-normalized before matching to prevent fullwidth-space/homoglyph bypass (e.g. git　push with U+3000)
• VMEM-0024 patch (RPAT-v3.5.1-2) — Refactored _Args class, pinned pip dependencies
• OLLAM-0023 patch (RPAT-v3.5.1-1) — Normalized scoring input, deduplicated level helper, hardened exclude patterns, added api_base warning
• Worktree teardown (CTDF-0025) — Task branch is now merged into local develop before worktree removal

Documentation

All 9 documentation sections updated to reflect v3.5.1 changes.

---

Install: /plugin install ctdf@dnviti-claude-task-development-framework
Update: /plugin update ctdf@dnviti-claude-task-development-framework

v3.5.0

What's New in v3.5.0

Added

• Vector Memory Layer (VMEM-0017) — Embedded vector database with LanceDB, AST-aware chunking, and semantic search
• MCP Server (VMEM-0018) — Model Context Protocol server for platform-agnostic vector memory access
• Multi-Agent Memory Consistency (VMEM-0019) — Advisory locking, conflict resolution, and garbage collection for shared vector stores
• Multiplatform Hardening (XPLAT-0020) — Cross-platform support for Windows, Linux, and macOS with CI matrix testing
• Social Media Announcer (REL-0021) — Post-release announcements to Bluesky, Mastodon, Discord, Slack, and clipboard
• Ollama Integration (OLLAM-0022) — Hardware-aware local model setup with orchestrated task offloading

Full Changelog: v3.4.6...v3.5.0

v3.4.6

[3.4.6] - 2026-03-17

Added

• Universal platform abstraction layer for multi-agent compatibility (XPLAT-0013) (#57)
• Cross-platform config generator for OpenCode, OpenClaw, Cursor, and AGENTS.md (XPLAT-0015) (#58)
• ZIP-based portable distribution for agentless manual installation (DIST-0014) (#59)
• ccpkg-compatible packaging for standardized cross-tool distribution (DIST-0016) (#60)
• show_generated_footer toggle to project config (CFG-0012) (#41)