tesseract CLAUDE.md

Tesseract — Claude Code Plugin Marketplace

Project Overview

This is a Claude Code plugin marketplace containing multiple independent plugins. Each plugin lives in its own top-level directory and can be installed separately.

Repository Structure

tesseract/
├── .claude-plugin/marketplace.json   # Marketplace registry — lists all plugins
├── infra-review/                     # Plugin: Terraform/Atmos review agents
│   ├── .claude-plugin/plugin.json
│   ├── agents/                       # Subagent definitions (markdown)
│   ├── commands/                     # Slash commands (markdown)
│   ├── hooks/                        # SessionStart + PostToolUse hooks
│   ├── scripts/                      # Helper scripts for hooks
│   └── skills/                       # Auto-invoked skills (SKILL.md)
├── clickup-sprint-planner/           # Plugin: ClickUp sprint planning MCP server
│   ├── .claude-plugin/plugin.json
│   ├── .mcp.json                     # MCP server configuration
│   ├── commands/                     # Slash commands
│   ├── server/                       # Python MCP server (mcp[cli] + httpx)
│   ├── skills/                       # Auto-invoked skills
│   └── examples/                     # Example config files
└── dev-workflow/                     # Plugin: Research + TDD skills
    ├── .claude-plugin/plugin.json
    └── skills/                       # Auto-invoked skills

Key Conventions

• Plugin isolation: Each plugin is self-contained. Changes to one plugin must not break others.
• Versioning: For relative-path plugins (all plugins in this repo), the version lives only in .claude-plugin/marketplace.json — NOT in plugin.json. The plugin.json manifest wins silently and can cause the marketplace version to be ignored. Additionally, bump pyproject.toml if the plugin has a Python server.
  • Reference: https://code.claude.com/docs/en/plugin-marketplaces#version-resolution-and-release-channels
  • Reference: https://code.claude.com/docs/en/plugins
• Commands are markdown files in commands/ — they define slash commands users can invoke.
• Skills are markdown files at skills/<name>/SKILL.md — they are auto-invoked by Claude when relevant.
• Agents are markdown files in agents/ — they define specialized subagents.
• Hooks use hooks.json to register shell scripts for Claude Code lifecycle events.
• MCP servers are configured in .mcp.json at the plugin root.
• Doc-authoring skills (any skill that writes a doc to docs/shield/) MUST apply the shield:writing-style skill to their author-written prose before writing — see shield/skills/general/writing-style/SKILL.md. This keeps generated docs simple, clear, and concise. It applies to prose only, never to rendered/marker-wrapped sections or JSON sidecars.