# Update Plugins

> Update all installed plugin marketplaces and their plugins concurrently. Use when the user asks to update plugins, check for updates, or refresh their plugin installations.

- Skill: `necturalabs/update-plugins` (Agent Skill)
- Install (CLI): `npx skillmds@latest add necturalabs/update-plugins`
- Raw SKILL.md: https://api.skillmd.com/api/skills/necturalabs/update-plugins/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: NecturaLabs (https://skillmd.com/u/necturalabs)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/necturalabs/update-plugins

---


# Update Plugins

Update all installed plugin marketplaces concurrently, then reinstall plugins that have newer versions available.

## Procedure

### Step 1: Discover Installed Marketplaces

Run `claude plugin marketplace list` to get all registered marketplaces. Parse the output to extract marketplace names.

### Step 2: Update All Marketplaces Concurrently

Use the `superpowers:dispatching-parallel-agents` pattern — launch one subagent per marketplace, each running:

```
claude plugin marketplace update <marketplace-name>
```

All agents run in parallel. Collect results from each.

### Step 3: Update Installed Plugins

After marketplace updates complete, run `claude plugin list` to see installed plugins. For each installed plugin, update it:

```
claude plugin update <plugin-name>@<marketplace-name>
```

Run these concurrently as well. This uses version comparison — plugins already at the latest version are skipped.

### Step 4: Report Results

Present a summary table:

| Marketplace | Status |
|-------------|--------|
| name | Updated / Already up to date / Failed: reason |

| Plugin | Version | Status |
|--------|---------|--------|
| name@marketplace | vX.Y.Z | Updated / Already current / Failed: reason |

## Error Handling

- If a marketplace update fails, report it but continue with the others
- If a plugin reinstall fails, report it but continue with the others
- Never abort the entire operation due to a single failure