Successfully installing the OpenClaw plugin does not necessarily mean new tools will appear. First, let's look at what the plugin manifest actually states: some plugins only provide skills, some only provide prompts or themes, and some register extensions and tools. Mistakenly thinking "installation successful" as "tool loaded" is the most common mistake when troubleshooting plugin issues.
First, clarify what plugins offer
If you have a writing skill, it might only affect the Agent's way of doing things, without adding a button-based tool; If it's a browser, file, message, or external API extension, you should see the corresponding capabilities in the tool list. There is no declaration function in manifest, so no matter how many times you restart, it won't appear out of thin air.
Then check if it connects during runtime
The plugin declares extensions, but you also need to check whether the current agent's runtime, tools.profile, and permission policies allow loading. For example, a tool requires system commands, but the current profile only supports messaging; Or the plugin directory permissions are abnormal and are skipped by security checks; Or the plugin dependencies may not be fully installed, causing manifests to be readable and the actual extension failing to start.
It is recommended to check in this order: confirm the plugin directory exists, read the manifest, check the OpenClaw status and logs, and then look at the current Agent's tool profile. Don't delete reinstalls right away, especially plugins installed from ClawHub or third-party sources, as reinstalls can mask real permissions and dependency issues.
If it's a team environment, before the plugin goes live, you also need to check which files it reads, whether it can connect to the internet, whether it requires keys, and whether it registers high-risk commands. The closer the tool is to the system layer, the more it should be tested on a single agent rather than directly applying to all entry points.
