Skills that teach an AI agent to work with MyList — the lists of things people love — through the mylist command line.

The skill assumes the CLI is installed and signed in:

npm install -g @mylist-world/cli
mylist auth login        # opens mylist.world to authorize this machine

A hosted agent — ChatGPT, Claude on claude.ai or Cowork, anything that cannot run a command — connects to MyList's MCP server instead of the CLI:

https://mylist.world/api/mcp

Add it as a custom connector or remote MCP server. The first connection opens MyList to sign in and approve the app; nothing to paste. The tools mirror the CLI (list_lists, get_list, create_item, add_to_list, search, fetch, …) and every connected app shows up under Settings › Apps & command line, where it can be disconnected. This skill is not needed there: the server describes its own tools.

Claude Code — as a plugin from this repo's marketplace:

/plugin marketplace add rivertwilight/mylist-skills
/plugin install mylist@mylist-skills

or, from a shell:

claude plugin marketplace add rivertwilight/mylist-skills
claude plugin install mylist@mylist-skills

Claude Code, without the plugin system — copy the skill folder to ~/.claude/skills/mylist-cli/ (every project) or .claude/skills/mylist-cli/ (one project).

Any other agent that reads SKILL.md folders (the Agent Skills layout) — copy skills/mylist-cli/ into wherever that agent loads skills from. The skill needs nothing but a shell with mylist on the path.

skills/mylist-cli/
├── SKILL.md                  what the agent reads when the skill triggers
└── references/
    ├── commands.md           every command, flag, exit code and JSON shape
    └── recipes.md            multi-step jobs: build a list from saved things, bulk tag, export, dedupe
.claude-plugin/
├── marketplace.json          lets Claude Code add this repo as a marketplace
└── plugin.json               the plugin the marketplace publishes (the repo root)

The CLI lives in the MyList monorepo under cli/. When a command or flag changes there, references/commands.md here is the file to update — it is what agents plan against.