Essay says Claude Mods turn the agent harness toward the user
Original title: Claude Mods: Flipping the Harness Inside Out
Anthropic ships six built-in Claude Code mods, and the author released an open-source ModSmith pack.
A Latent Space essay describes Claude Mods as plugins of TypeScript or JavaScript that change how Claude Code renders and handles events in the CLI and desktop app. Anthropic lists six built-in mods, including a side agent called You Should Know that is off by default, and Claude Code can generate mods from plain-language requests. The author highlights a personal Next Steps Supervisor mod in the open-source ModSmith plugin. The piece also points to OpenAI's ChatGPT Intelligent UI, which it says shipped with GPT-6 on October 7, and to Cursor approval flows.
Unconfirmed: The essay is the author's product commentary, including the claim that OpenAI shipped GPT-6 on October 7.
- Built-in mods include agents-md, diff, plugin-authoring, sec-default, telemetry, and you-should-know.
- You Should Know flags material actions buried in the agent trace.
- Next Steps Supervisor runs after file edits, commands, or eight or more tool calls.
- sec-default is described as guarding organization-managed settings from user-installed mods.
- Products
- Claude CodeChatGPT
- AI models
- GPT-6
From the original article
9 min read at latent.space
Claude Mods is an early form of the Attention Interface. It flips the agent harness paradigm from the harness facing the machine, to the harness facing the human.
By Dan McAteer · latent.space
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.