JeriCommerce Claude Code Plugins

Official plugin marketplace for Claude Code by JeriCommerce.

Installation

Add this marketplace to your Claude Code:

/plugin marketplace add JeriCommerce/claude-plugins

Then install individual plugins:

/plugin install loyalty-specialist@jericommerce-plugins

Available Plugins

Plugin	Description	Version
loyalty-specialist	Jeri - Generate professional PDF loyalty program performance reports by querying Metabase and analyzing program data	202603.10.1
bug-reporter	Jeri - Report bugs and customer issues to Linear through guided conversational flow with screenshots/video support	202602.19.0
qa-specialist	Jeri - QA testing assistant — verify features, test endpoints, inspect logs, and document findings in Linear	202602.21.0
feature-request	Jeri - Capture customer feature requests as user stories in Linear with customer association	202603.11.0
loyalty-template-creator	Jeri - Create CDN loyalty page templates for Shopify theme extensions by replicating a reference website's style	202603.10.0
niche-research	Jeri - Generate pSEO niche research reports from Ahrefs data, ready to copy into the JeriCommerce blog repo	202603.12.0
optimize-skills	Audit and restructure Claude Code skills into modular folder structures for better consistency and maintainability	202603.21.0

Updating

To get the latest plugins after changes are pushed:

/plugin marketplace update

---

Creating a New Plugin

1. Directory structure

Create a new folder under plugins/ with this structure:

plugins/my-plugin/
├── .claude-plugin/
│   └── plugin.json        # Required: plugin metadata
├── skills/                 # Optional: auto-invoked by Claude when relevant
│   └── my-skill/
│       ├── SKILL.md        # Skill definition (frontmatter + instructions)
│       └── references/     # Optional: supporting files the skill can read
│           └── data.md
├── commands/               # Optional: user-invoked via /my-plugin:command-name
│   └── my-command.md
├── agents/                 # Optional: specialized subagents with isolated context
│   └── my-agent.md
└── README.md               # Recommended: plugin documentation

All directories are optional except .claude-plugin/. Include only what your plugin needs.

2. Plugin manifest (plugin.json)

Create .claude-plugin/plugin.json with your plugin's metadata:

{
  "name": "my-plugin",
  "description": "Short description of what the plugin does",
  "version": "1.0.0",
  "author": {
    "name": "JeriCommerce"
  },
  "keywords": ["relevant", "search", "terms"]
}

Field	Required	Description
name	Yes	Unique identifier in kebab-case. Becomes the namespace for commands (/my-plugin:command)
description	No	Shown in the plugin manager when browsing plugins
version	No	Semantic versioning (MAJOR.MINOR.PATCH). Bump this on every change or users won't see updates
author	No	Attribution. { "name": "...", "email": "..." }
keywords	No	Tags for plugin discovery

3. Skills (auto-invoked by Claude)

Skills are loaded automatically when Claude determines they're relevant to the current task. Create a folder inside skills/ with a SKILL.md file.

skills/my-skill/SKILL.md:

---
name: my-skill
description: >
  Describe WHEN Claude should use this skill. Include trigger phrases
  and scenarios. Example: "Use when the user asks about X, works on Y,
  or mentions Z."
---

# My Skill

Instructions for Claude when this skill is activated.

## Reference Files

You can point Claude to supporting files in the same directory:

Read `${CLAUDE_PLUGIN_ROOT}/skills/my-skill/references/data.md` for details.

Frontmatter fields:

Field	Required	Description
name	Yes	Skill identifier
description	Yes	Tells Claude when to activate the skill. Be specific with trigger phrases
version	No	Skill version
disable-model-invocation	No	Set to true if only the user should invoke it (not Claude automatically)

Tips:

• Use ${CLAUDE_PLUGIN_ROOT} to reference files within your plugin directory
• Put large reference data in references/ subdirectory to keep SKILL.md focused
• The description field is critical — Claude reads it to decide when to use the skill

4. Commands (user-invoked via slash command)

Commands are triggered explicitly by the user typing /plugin-name:command-name. Create .md files inside commands/.

commands/my-command.md:

---
description: Short description shown in /help
argument-hint: [required-arg] ["optional arg"]
---

Parse `$ARGUMENTS` to extract:
- **First argument** (`$1`): description
- **Second argument** (optional): description

## Steps

1. First, read the skill file at `${CLAUDE_PLUGIN_ROOT}/skills/my-skill/SKILL.md`
2. Do something with the arguments
3. Produce output

Frontmatter fields:

Field	Required	Description
description	Yes	Shown when the user runs /help
argument-hint	No	Usage hint shown to the user

Variables available:

• $ARGUMENTS — Full text after the command name
• $1, $2, ... — Positional arguments
• ${CLAUDE_PLUGIN_ROOT} — Absolute path to the plugin directory

5. Agents (specialized subagents)

Agents are specialized AI assistants that run in their own context window. Claude can delegate tasks to them automatically, or the user can request them explicitly. Create .md files inside agents/.

agents/my-agent.md:

---
name: my-agent
description: >
  Describe WHEN Claude should delegate to this agent. Be specific.
  Example: "Use proactively when the user needs to analyze data,
  debug queries, or investigate production issues."
tools: Read, Grep, Glob, Bash
model: sonnet
---

You are a specialist in X. When invoked:

1. Understand the task
2. Gather context by reading relevant files
3. Perform the analysis
4. Return a clear summary of findings

Key rules:
- Always do A before B
- Use tool X for Y
- Format output as Z

Frontmatter fields:

Field	Required	Description
name	Yes	Unique identifier (lowercase, hyphens)
description	Yes	Tells Claude when to delegate. Include "use proactively" for auto-delegation
tools	No	Allowed tools (inherits all if omitted). Options: Read, Edit, Write, Bash, Grep, Glob, Task, MCP tools
disallowedTools	No	Tools to explicitly deny
model	No	sonnet, opus, haiku, or inherit (default)
permissionMode	No	default, acceptEdits, dontAsk, bypassPermissions, plan
maxTurns	No	Max agentic turns before stopping
skills	No	Skills to preload into the agent's context
mcpServers	No	MCP servers available to this agent
hooks	No	Lifecycle hooks scoped to this agent
memory	No	Persistent memory scope: user, project, or local

Key differences from Skills:

	Skills	Agents
Context	Runs in the main conversation	Runs in its own isolated context
Tools	Uses main conversation tools	Can have restricted/custom tool access
Model	Same as main conversation	Can use a different model
Output	Inline in conversation	Returns a summary to the main conversation
Best for	Domain knowledge, workflows	Isolated tasks, parallel research, restricted operations

Tips:

• Agents preserve the main conversation's context by running separately
• Use tools to restrict what the agent can do (e.g., read-only agents)
• Agents cannot spawn other agents
• Use skills field to preload domain knowledge into the agent
• MCP tools are referenced by their full name (e.g., mcp__metabase__run-native-query)

6. Register in the marketplace

Add your plugin entry to .claude-plugin/marketplace.json in the plugins array (see step 2 for version format):

{
  "name": "my-plugin",
  "source": "./plugins/my-plugin",
  "description": "Same or similar description as plugin.json",
  "version": "1.0.0",
  "author": {
    "name": "JeriCommerce"
  },
  "keywords": ["relevant", "terms"],
  "category": "development"
}

Field	Required	Description
name	Yes	Must match the name in plugin.json
source	Yes	Relative path to the plugin directory (always ./plugins/plugin-name)
description	No	Shown when browsing the marketplace
version	No	Should match plugin.json. Avoid setting it in both places — plugin.json always wins
category	No	For organization (e.g., development, analytics, devops)

7. Update the Available Plugins table

Add a row to the table in this README:

| [my-plugin](./plugins/my-plugin/) | Short description | 1.0.0 |

8. Commit and push

git add -A
git commit -m "feat: add my-plugin"
git push origin main

Users will see the new plugin after running /plugin marketplace update.

---

Quick Reference: Plugin Components

Component	Location	Invoked by	Use case
Skills	skills/name/SKILL.md	Claude (automatic)	Domain knowledge, context, workflows that Claude applies when relevant
Commands	commands/name.md	User (/plugin:name)	Explicit actions with arguments (generate reports, run analyses)
Agents	agents/name.md	Claude or user	Specialized subagents for complex tasks
Hooks	hooks/hooks.json	Events (automatic)	React to events like file saves, tool use, etc.
MCP Servers	.mcp.json	Automatic	Connect to external tools and APIs

---

Example: Minimal Plugin

The simplest possible plugin with one skill:

plugins/my-helper/
├── .claude-plugin/
│   └── plugin.json
└── skills/
    └── helper/
        └── SKILL.md

plugin.json:

{
  "name": "my-helper",
  "description": "Provides context about X",
  "version": "1.0.0"
}

SKILL.md:

---
name: helper
description: Use when the user asks about X or works on X-related code.
---

When helping with X, follow these rules:
1. Always do A
2. Never do B
3. Prefer C over D

marketplace.json entry:

{
  "name": "my-helper",
  "source": "./plugins/my-helper",
  "description": "Provides context about X",
  "version": "1.0.0"
}

---

License

Proprietary — JeriCommerce