Full Tooling Inventory

Global Commands (~/.claude/commands/)

Command	Model	Purpose
/requirement-intake	sonnet	NL requirement → REQ spec → route to Obsidian dept folder
/obsidian-doc-sync	sonnet	Post-feature: sync code changes → Obsidian vault
/code-review	sonnet	Review uncommitted changes against universal checklist
/auto-commit	haiku	Stage + conventional commit with AI message
/quick-search	haiku	Fast codebase search via Explore subagent

Global Agents (~/.claude/agents/)

Agent	Model	Purpose
code-reviewer	sonnet	Project-aware code review (reads CLAUDE.md first)
cross-project-scout	sonnet	Cross-repo pattern search, utility detection, dep audit

Project-Local: project_root only (.claude/commands/ + .claude/skills/)

The only project with a real delivery pipeline:

/plan  →  /build  →  /update-docs  →  /auto-commit
  ↕           ↕           ↕
planner    developer    updater      (skills backing each command)
sonnet      opus        sonnet

Command/Skill	Model	Purpose
/plan + planner	sonnet	Read design docs → structured implementation plan
/build + developer	opus	Implement plan → run pytest → implementation summary
/update-docs + updater	sonnet	Diff code → update ERD/architecture/decisions docs
/code-review + code-reviewer	sonnet	ICP-specific review: PEP8, type hints, SQL safety
/quick-search	haiku	ICP-scoped fast search
/auto-commit	haiku	Commit
/weekly-report	sonnet	Git history → weekly summary

Plugin Skills (~/.claude/plugins/)

superpowers (always active)

Skill	Trigger
brainstorming	Before any creative/feature work
writing-plans	Create implementation plans
executing-plans	Execute a written plan
subagent-driven-development	Orchestrate spec/implement/review subagents
dispatching-parallel-agents	2+ independent tasks in parallel
using-git-worktrees	Isolated branch work
systematic-debugging	Any bug/test failure
test-driven-development	Before writing implementation code
verification-before-completion	Before claiming work done
finishing-a-development-branch	Merge/PR decision
requesting-code-review	After feature completion
receiving-code-review	When handling review feedback
writing-skills	Creating new skills

claude-reflect (self-learning)

Command	Purpose
/reflect	Process queued learnings → update CLAUDE.md
/reflect-skills	Discover skill candidates from session patterns
/view-queue	See pending learnings

Hooks fire on every session: UserPromptSubmit, PostToolUse, PreCompact — captures corrections and patterns into learnings-queue.json.

ralph-loop

/ralph-loop — continuous Claude loop for iterative tasks until completion.

voltagent-infra

14 DevOps/cloud specialist subagents (Azure, K8s, Terraform, SRE, etc.).

Pattern Map: What You Actually Use

INTAKE                 PLANNING              DELIVERY              DOCS/LEARNING
─────────────────────────────────────────────────────────────────────────────────
/requirement-intake    brainstorming skill   /build (opus)         /update-docs
  → Obsidian spec      writing-plans skill   /code-review          /obsidian-doc-sync
                       /plan (sonnet)        /auto-commit          /reflect
                                             ralph-loop            cross-project-scout