v1.2.1

Fixed

• Tests no longer depend on the host machine's init.defaultBranch git config — test packages isolate from global/system git config via TestMain (xpo-3b2ec1)
• Mergeability check no longer falsely blocks on uncommitted working directory changes — now uses git merge-tree to compare committed refs only (xpo-fad57f)
• Merge view disables the merge button and shows "No commits to merge" when there are zero commits ahead

Added

• MergeView Files tab scope toggle: switch between "All changes" (branch vs main) and "Uncommitted" (working tree vs HEAD) with 3-second auto-refresh polling; includes synthetic diffs for untracked files (xpo-6211f5)
• Amber file icon highlighting in MergeView file tree for uncommitted/untracked worktree files; mergeability API now returns dirty_files array (xpo-110b09)
• Multi-harness init: xpo init is now config-only; new xpo init mcp writes per-harness MCP server config, xpo init skill installs workflow skills and agent instruction files — each concern is independently runnable (xpo-ceb434)
• Agent harness support for Claude Code, GitHub Copilot, Cursor, OpenAI Codex, and OpenCode — per-harness MCP config formats (JSON with mcpServers/mcp keys, TOML, and OpenCode's local-array style) and skill directories (xpo-ceb434)
• Global skill installation (xpo init skill --global): canonical location at ~/.config/xpo/skills/ with symlinks into each harness's skill directory; macOS/Linux only (xpo-ceb434)
• xpo init skill interactive flow: detects installed agents, prompts for global/local scope, respects existing agent instruction files (append without --force, replace xpo section with --force) (xpo-ceb434)
• xpo doctor reorganized into three sections — Project Configuration, MCP Configuration, Agent Integration — each with its fix command (xpo-ceb434)
• xpo init prints actionable next-step hints when MCP or skills are not yet configured (xpo-ceb434)
• 20 new scenario tests covering init/mcp/skill clean, idempotent, force, and cross-format workflows (xpo-ceb434)

Changed

• Skill upgraded to v3.1: restructured workflow steps (rationale tool, explicit gates, handoff self-review, issue labels with primary/secondary system, backlog review section, CANCELED/DUPLICATE statuses) (xpo-dd97a5)
• Skill directory renamed from xpo-workflow to xpo to match reference skill (xpo-dd97a5)
• Colored CLI output: green ✓ success, yellow ! warnings, red ✗ errors via lipgloss (xpo-dd97a5)
• xpo init prompts for issue ID prefix interactively, shows quick health check with MCP/skill status, single call-to-action for xpo doctor (xpo-dd97a5)
• xpo doctor summary line ("7 passed, 4 warnings"), one fix hint per section, tighter formatting (xpo-dd97a5)
• Root command description: "The git-native engineering system for human-AI teams" (xpo-dd97a5)
• Agent registry trimmed from 10 to 6 harnesses: Generic Agent, Claude Code, GitHub Copilot, Cursor, Codex, OpenCode — removed Windsurf, Gemini, Cline, Roo Code, Aider, Continue (xpo-ceb434)
• PersistentPreRunE allowlist uses parent-chain check so init subcommands work without config (xpo-ceb434)