Anthropic has released Claude Mods. You can now hook into the harness and UI and change them. There are so many possibilities that it’s a bit hard to choose.
- You can synchronize parallel work across several sessions using a shared file of edits + messages between sessions.
- You can automatically check the agent’s answers for cognitive drift and ask it to rewrite them.
- You can automatically turn remarks beginning with “Ошибка!…” into a task to update the rules, with a diff and confirmation.
- You can catch tokens, keys, and secrets in command output and replace them with “[hidden]” before they are recorded in the conversation.
- You can remove skills, agents, and tools that aren’t needed for the current task from the list the model sees.
- You can show the full list of tasks and action points for the current session and their status in a panel.
- You can implement a full-fledged critic through external systems.
- You can show current alerts and buttons to start investigations in a side panel.
https://code.claude.com/docs/en/plugins/mods/reference#mods-api-methods