ark-skills

Comprehensive multi-agent code review that orchestrates Claude Code agents (code-reviewer, code-architect, test-coverage-checker, silent-failure-hunter, test-analyzer) into a unified review. Includes an epic review mode that pulls Obsidian TaskNotes epics, stories, and session logs to cross-reference planned work against actual code changes. Also includes a plan review mode that cross-references code against implementation plans and design specs. Includes a full review mode (--full) that combines thorough code analysis with auto-detected epic context. Use this skill whenever the user mentions "code review", "review my changes", "review before deploy", "check my code", "ark review", "/ark-code-review", "pre-deploy review", "review this branch", "review these files", "review this epic", "does the code match the plan", "review against plan", "review against spec", "check plan completion", "did I implement the plan", "full review", "complete review", "review everything", or any request to get feedback on code quality. Also triggers on "what did I break", "is this safe to deploy", "sanity check", or "second opinion on this code". Do NOT use for: PR reviews on GitHub (use code-review:code-review directly), codebase maintenance/cleanup (use codebase-maintenance), or general questions about the codebase.

Load recent + relevant project context before any /ark-workflow chain. Runs as step 0 of every chain. Queries /notebooklm-vault (if set up), /wiki-query (if set up), and /ark-tasknotes (if set up) to emit a single Context Brief. Triggers on "warm up", "context brief", "load context", "start fresh on this project". Also invoked automatically as chain step 0 by /ark-workflow.

Diagnostic check for Ark ecosystem health — plugins, CLAUDE.md, vault, integrations

Interactive setup wizard — greenfield, vault migration, partial repair. Absorbs /wiki-setup.

Agent-driven task creation and management via tasknotes MCP or direct markdown write