v0.1.9

New

• Codex skill supported.
• GitHub Copilot CLI skill supported.
• Gemini CLI skill supported.
• Dry-run-first refinement: /deeprefine now stops after the review report and does NOT touch graphify-out/graph.json until the user explicitly approves. The agent loop is: history sync-memory → load pending → refine → review → wait → apply → loop finish.
• Evidence-aware action review.

Next

• Do coverage-based greedy query selection.
• Adapt to more agent frameworks (Claude Code, Cline, etc.).