ControlScope, a research system for LLM agents, compares three ways a running workflow can be revised from the same public execution state: continuing generated code, editing the next tool call's data arguments, and replacing the unfinished workflow. The work frames these as nested permissions that separate available repairs from the actions an agent is actually allowed to take.
How much of a running workflow should a language model agent revise? ControlScope compares continuing generated code, editing the next tool call's data arguments, and replacing the unfinished workflow from the same public execution state. The nested permissions separate available repairs from the ac
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.