Scraped the mod zone: Anthropic has added mods to Claude Code—just a few lines of TypeScript lets you change behavior, redraw the interface, or even replace built-in functionality entirely; with /plugin, you can even have Claude write it for you on the spot, and hot-reload without restarting.
The capabilities are broader than the old hooks: you can modify prompts, intercept tool calls, manage permissions, or even scrub secrets out of the output before anything is shown on the screen. In the official demo there are three—Token Weather uses context to estimate token usage, Blast Radius lists what will be deleted before removing files so you can confirm, and Replay Theater runs a round to show which diffs were changed. The roadmap is to turn more built-in features into mods, shrinking the core further.
The trade-offs are stated just as plainly: mods aren’t sandboxed—permissions are equivalent to Claude Code itself, and should only be installed from sources you trust. Team/Enterprise will load sec-default first to block dangerous overrides. As freedom increases, the risk is handed back to the installer.
#ClaudeCode #Anthropic #AI
The capabilities are broader than the old hooks: you can modify prompts, intercept tool calls, manage permissions, or even scrub secrets out of the output before anything is shown on the screen. In the official demo there are three—Token Weather uses context to estimate token usage, Blast Radius lists what will be deleted before removing files so you can confirm, and Replay Theater runs a round to show which diffs were changed. The roadmap is to turn more built-in features into mods, shrinking the core further.
The trade-offs are stated just as plainly: mods aren’t sandboxed—permissions are equivalent to Claude Code itself, and should only be installed from sources you trust. Team/Enterprise will load sec-default first to block dangerous overrides. As freedom increases, the risk is handed back to the installer.
#ClaudeCode #Anthropic #AI
