ClosedLoop.ai
MechanismsPlugins

platform plugin

Claude Code expert guidance, prompt engineering, mermaid diagrams, artifact upload, and skill scaffolding.

The platform plugin is a standalone reference plugin that ships five auto-triggered skills. It has no slash commands — skills activate based on context.

Skills (5)

claude-code-expert

Quick reference for Claude Code:

  • YAML frontmatter patterns for agents, commands, and skills
  • 10 hook event types
  • 14 hook recipes
  • Decision framework: when to build a skill vs an agent vs a command vs a plugin
  • Style requirements (imperative voice, not second-person)

References include: agent-patterns.md, command-patterns.md, configuration.md, hook-recipes.md, skill-triggers.md, and sub-agents-config.schema.json.

context-engineering

9 techniques ranked by effectiveness:

  1. Be clear and direct
  2. Multishot examples
  3. Chain of thought
  4. XML tags
  5. Role prompting
  6. Prefill response
  7. Chain prompts
  8. Long context tips (200K-token patterns: data at top, XML doc structure, grounding quotes)
  9. Extended thinking (for STEM and constraint optimization)

mermaid-visualizer

6 diagram types with a syntax reference listing prohibited symbols and safe alternatives. Used to generate architecture and flow diagrams inside docs or plans.

upload-artifact

Two modes for getting artifacts into ClosedLoop.ai:

Script mode — runs scripts/upload_artifact.py under uv as a direct Streamable-HTTP MCP client. The file content does not enter the conversation context.

Required environment:

  • CLOSEDLOOP_API_KEY
  • NEXT_PUBLIC_MCP_SERVER_URL

Flags:

--url --api-key --list-projects --file --title
--type PRD|IMPLEMENTATION_PLAN|TEMPLATE
--project-id --workstream-id --artifact-id --verify

MCP fallback — uses mcp__closedloop__create-artifact or mcp__closedloop__create-artifact-version. Loads file content into context.

claude-creator

Scaffolds a new skill with SKILL.md, references/, scripts/, and assets/. Ships:

  • init_skill.py — scaffold
  • package_skill.py — package for distribution
  • quick_validate.py — local validation

Why these skills ship together

They are the toolbelt for working inside ClosedLoop.ai:

  • claude-code-expert teaches you the framework.
  • context-engineering teaches you how to prompt.
  • mermaid-visualizer teaches you to diagram.
  • upload-artifact lets you push work back into the control plane.
  • claude-creator lets you extend the system.

The platform plugin is the recommended first install for anyone learning how to build inside the ClosedLoop.ai ecosystem.

On this page