Plugin Ecosystem

Overview of the 10-plugin architecture — what each plugin does and how they work together.

New

"I Want To..." Guide

I want to...Use this pluginKey command
Brainstorm an ideasuperpowers/brainstorm
Write a spec or PRDspec-kit/speckit-specify
Review architecturepensive/architecture-review
Plan implementationsuperpowers/write-plan
Implement with TDDsuperpowers/execute-plan
Review codepensive/full-review
Commit changessanctum/commit-msg
Prepare a PRsanctum/prepare-pr
Work on an issuesanctum/do-issue
Debug a problemsuperpowerssystematic-debugging skill
Catch up on contextimbue/catchup
Optimize context usageconserve/optimize-context
Create a new skillabstract/create-skill
Check Ontopix patternsontopix/check-compliance
Research Ontopix docsontopix/search-docs
Run iterative loopsralph-wiggum/ralph-loop
Manage all pluginsleyline/update-all-plugins

Architecture

The plugin ecosystem follows a three-layer model:

Layer 3: ontopix plugin       → Live Ontopix patterns via MCP
Layer 2: sanctum/pensive/...  → Generic dev workflows (commits, reviews, TDD)
Layer 1: Claude Code          → Core runtime

Layer 1 — Claude Code is the core runtime. It provides the AI assistant, tool use, and the plugin system itself.

Layer 2 — Generic Plugins add development workflows that work in any project:

  • superpowers provides the core development loop: brainstorming, planning, implementing (TDD), and debugging.
  • sanctum handles git operations: commits, PRs, documentation updates, and dependency management.
  • pensive provides code review from multiple angles: bugs, architecture, tests, and domain-specific reviews.
  • imbue adds workflow methodologies: evidence gathering, structured reviews, and scope guarding.
  • abstract enables meta-operations: creating new skills, evaluating plugins, and managing the ecosystem.
  • conserve optimizes resource usage: token conservation, bloat detection, and context management.
  • leyline provides infrastructure building blocks consumed by other plugins: auth patterns, error handling, and git platform detection.
  • spec-kit adds spec-driven development: from specification to implementation with structured workflows.
  • ralph-wiggum enables iterative refinement loops using the technique from Anthropic.

Layer 3 — Ontopix Plugin connects Claude Code to your team's engineering patterns via MCP, making the AI aware of your conventions.

Plugin Comparison

PluginPurposeSkillsCommandsAgents
superpowersCore dev workflows1431
sanctumGit and workspace ops14189
pensiveCode review toolkit12124
imbueWorkflow methodologies1031
abstractMeta-skills infrastructure11185
conserveResource optimization1165
leylineInfrastructure building blocks1420
spec-kitSpec-driven development393
ralph-wiggumIterative loops030
ontopixOntopix patterns300
Total927428

Plugin Deep-Dives

  • superpowers — Core skills: TDD, debugging, brainstorming, planning
  • sanctum — Git workflows: commits, PRs, docs, dependencies
  • pensive — Code review toolkit: bugs, architecture, tests
  • imbue — Workflow methodologies: evidence, structured reviews, scope guard
  • abstract — Meta-skills: skill authoring, plugin validation, hooks
  • conserve — Resource optimization: tokens, context, bloat detection
  • leyline — Infrastructure: auth, errors, git platform, testing standards
  • spec-kit — Spec-driven development: specify, plan, implement
  • ralph-wiggum — Iterative loop technique from Anthropic

See Also