Reflexive-Claude-Code

Detects project languages and monorepo state, runs language-appropriate static analysis (dependency graph, complexity, duplication, semantic patterns), and produces a refactor map ranking hotspots. Use when user invokes /aref or explicitly asks to analyze a codebase for refactoring.

Executes a refactor plan phase-by-phase on a dedicated branch with per-phase commits and mandatory reviewer checkpoints. Use when characterization-tests scaffold is complete and plan has phases ready to execute.

Writes AGENTS.md per subproject, archives run artifacts, and suggests rcc handoff conditionally. Use when verifying-refactors passes (PASS or PASS-WITH-WEAK-TESTS).

Converts a refactor map into a phased plan using parallel-change, branch-by-abstraction, or strangler fig patterns. Use when user has approved the refactor map from analyzing-codebases.

Adds golden/snapshot tests to untested hotspot modules before refactoring. Use when refactor plan marks any phase with characterization_test.status=must-scaffold.

Validates hard structural rules (no cycles, file/fn line caps, cognitive/cyclomatic complexity) and runs mutation testing on touched modules. Use when applying-refactors has completed all phases on the refactor branch.

Validates component-type choices for agent system work, classifying knowledge as CLAUDE.md vs rule vs skill vs agent vs hook and checking for conflicts. Use when starting any skill/agent/rule workflow to validate approach. Use when classifying knowledge type. Use when checking for component conflicts.

Scans project structure and agent system components, runs 11-category weakness analysis, and produces actionable restructuring recommendations. Use when analyzing an existing agent system for weaknesses. Use when user says "analyze agent system", "check agent setup", "audit agent config". Use when called by migrating-agent-systems or refactoring-skills.

Builds agent system components from a confirmed component plan. Use when executing a component plan to build agent system elements. Use when called by planning-agent-systems after plan is confirmed. Use when user says "apply agent plan", "build agent system".

Explores user workflows through targeted questions about pipeline modes, pain points, routine tasks, and human intervention points to inform agent system design. Use when exploring workflows after analysis. Use when user says "explore workflows", "brainstorm workflows", "what should I automate". Use when called by analyzing-agent-systems.

Scaffolds a new Claude Code plugin package with manifest, skills, agents, and marketplace entry. Use when creating a new Claude Code plugin package. Use when user says "create plugin", "new plugin", "scaffold plugin", "plugin template".

Fans out work across multiple agents — same prompt to many for Best-of-N, different prompts for parallel exploration, or a scout before committing. Use when user says "parallel agents", "fan out", "best of N", "scout", "race", "vote", "spawn workers".

Fetches the current official Claude Code component spec from code.claude.com, returning verbatim excerpts so designs never drift from stale memory. Use PROACTIVELY before writing or modifying any Claude Code component (subagent, skill, hook, rule, slash command, plugin manifest, settings.json, MCP config). Use when user asks 'what is the official spec for X', 'is this the right way to do Y', 'check Anthropic docs for Z'. MUST BE USED at Task 0 of writing-subagents, writing-skills, writing-hooks, writing-rules, writing-claude-md before any design decision.

Optimizes a single existing skill through analysis, best-practice research, and targeted edits. Use when user says "improve skill", "optimize skill", "update skill", "skill needs work". Use when one specific skill is outdated or underperforming. NOT for multi-skill consolidation (use refactoring-skills instead).

Bootstraps a new project with scaffolding, tooling, and Claude Code agent system setup. Use when bootstrapping a new project from scratch. Use when user says "create project", "new project", "initialize project", "start new app".

Sets up or migrates a project's Claude Code agent system through the analyze, brainstorm, plan, apply pipeline. Use when user says "setup agent", "migrate agent system", "configure claude code", "add agent system".

Sets up or migrates a Claude Code plugin, routing to creating-plugins or the validating + refactoring pipeline. Use when user says "setup plugin", "migrate plugin", "plugin from scratch", "check my plugin", "turn this into a plugin", "package this as plugin".

Plans which agent system components to create or modify, producing a component plan for user confirmation. Use when called by brainstorming-workflows after workflow exploration is complete. Use when user asks to plan agent system components.

Cleans up and consolidates an agent system after creation or modification. Use when user says "cleanup agent system", "refactor agent setup", "clean up skills", "consolidate agents". Use when called by applying-agent-systems.

Refactors and audits Claude Code plugin packages against official best practices. Use when refactoring, migrating, or auditing a Claude Code plugin package. Use when user says "refactor plugin", "audit plugin", "plugin health check", "migrate plugin structure". Use when plugin structure drifts from official best practices.

Analyzes and refactors multiple skills across a project, consolidating overlaps, extracting shared conventions, and removing redundancy. Use when user says "refactor skills", "cleanup skills", "consolidate skills", "skill review". Use when skills overlap, are redundant, or need cross-skill consolidation. NOT for improving a single skill (use improving-skills instead).

Scans the session, suggests what is worth reflecting on, and lets the user pick which learnings to deepen. Use when completing significant work, when the user corrects your approach, when a bug is resolved, or when new patterns emerge. Use when user says "reflect", "what did we learn", "capture learnings".

Reviews agent system components with specialized reviewer subagents after creation or modification. Use when called by applying-agent-systems after all components are built. Use when user says "review agent system", "check quality".

Scans all SKILL.md, agent, rules, and plugin manifest files for frontmatter errors, broken links, orphaned files, and invalid variables, generating a Markdown report with targeted fixes. Use when auditing plugin structure after migration or refactoring. Use when user says "validate plugin", "audit skills", "check plugin structure", "run validation after refactor".

Creates or improves CLAUDE.md project configuration with token-efficient, actionable instructions. Use when creating CLAUDE.md, improving existing CLAUDE.md, or setting up project configuration. Use when user says "create CLAUDE.md", "setup project", "configure agent".

Creates Claude Code hooks that enforce code quality, static analysis, or workflow automation deterministically. Use when user says "add hook", "enforce linting", "add pre-commit check", "block bad code".

Creates scoped convention rules in .claude/rules/ that auto-inject into matching contexts. Use when adding project conventions or scoping guidelines. Use when user says 'add convention', 'scope guideline', 'add rule', 'create rule'.

Creates and improves Claude Code skills following the official Anthropic spec and house standards. Use when creating new skills, modifying existing skills, or improving skill quality. Use when user says "create a skill", "write a skill", "add capability". Use when a repeated pattern should become reusable.

Creates specialized Claude Code subagents in .claude/agents/ with minimal tools, clear triggers, and isolated context. Use when user says 'create agent', 'add reviewer', 'specialized agent', 'isolated context task'.