Agent skills for AI app builders

Use your agent to choose a category, prepare a submission, improve your listing and track competitors across the AI app stores.

Skills are instructions your agent loads for a task. Plugin Stack is OpenApps' collection for people building ChatGPT apps, Claude plugins and connectors, Cursor plugins and MCP servers.

Install

Any agent

Command

npx skills add theopenapps/plugin-stack

Run this in a terminal. The installer finds your agents and asks where to put the skills. Add --skill optimize-listing to install just that skill.

Inside a Claude Code session

Command

npx skills add theopenapps/plugin-stack -a claude-code

Name the agent when it runs the installer for you. Without this option, a non-interactive install writes to .agents/skills/, which Claude Code does not read.

Claude Code plugin

Command

/plugin marketplace add theopenapps/plugin-stack
/plugin install plugin-stack@theopenapps

Run these commands in Claude Code. This installs the collection as a plugin.

Codex plugin

Command

codex plugin marketplace add theopenapps/plugin-stack
codex plugin add plugin-stack@theopenapps

Run these commands in a terminal to add the collection to Codex.

You can also copy an individual skill folder from the repository into your agent's skills directory. Keep its reference files with it.

Choose a skill

plugin-stack

Read your project, find one useful starting point and choose what to do next.

Try asking: "We built the plugin. Now what?"

find-opening

Compare categories and competitors to find openings for a new app.

Try asking: "Which ChatGPT app categories have room for a new product?"

submit-plugin

Check submission readiness and prepare a kit for ChatGPT, Claude or Cursor.

Try asking: "Help me submit our MCP server to Claude's directory."

track-launch

Compare your review wait with published apps and check where your app is listed.

Try asking: "Our ChatGPT app has been in review for nine days. What should we check?"

design-widgets

Study widgets from live ChatGPT apps and review your own design.

Try asking: "Review our widget against examples from live ChatGPT apps."

optimize-listing

Review listing fields against competitors and draft copy within the store's limits.

Try asking: "How could we improve our Claude plugin's description?"

audit-discoverability

Check whether your listing and tools fit the prompts you want to be found for.

Try asking: "Why might an agent overlook our app for these prompts?"

watch-competitors

Compare ranks, featuring, listing edits and tool changes since your last check.

Try asking: "What changed for our competitors this week?"

Start with your project

Ask "We built the plugin. Now what?" The plugin-stack skill reads your project and the stores, then suggests next steps. With your approval, it saves the project context in .agents/plugin-stack.md. Each skill also works on its own.

Data and access

Terminal agents read OpenApps through the CLI, so you can use these skills without an MCP connection. Agents without a shell can use the MCP server or read the REST API. Public data needs no account; your account's plan determines the history and rows available.

The reports date their evidence and say when a limit leaves data out. Store rankings do not measure demand or revenue. The discoverability check simulates an agent's choice; it does not run a store search.

Installation and updates on GitHub · Plans and limits