Long-running tasks: a protocol for AI agents to work for hours or days
The discussion presents a `/goal` prompt protocol for continuous AI agent tasks amid the shift toward autonomous agents.
The /goal protocol organizes long tasks for AI agents that may work for hours or days. It defines a verifiable objective and a condition for ending the work. The proposal comes amid a shift from chatbots to autonomous agents that can use a computer, browser, and tools. OpenAI Dots is cited as an example of this model.
For users, the change is moving from isolated requests to a task tracked in stages. The agent analyzes, executes, tests, corrects, updates the record, and continues. Files such as goal.md, plan.md, and state.md preserve important information so work can continue even after the context is compacted. The guidance is to let the agent decide on low-risk, reversible technical actions, and involve a person when there is a real blockage, destructive risk, or business decision.
To get started, the objective should specify a result that can be checked and when the work should stop. The full protocol can live in AGENTS.md so it does not need to be repeated in every session; the /goal command is reserved for the specific objective. It is also important to track tests, corrections, and progress records. The project guide is available at https://inematds.github.io/execucao-longa/guia/.
The link opens the original content; some require a subscription.
Go further on INEMA.CLUB
More news
- 02/10NVIDIA launches an open platform to control AI agentsThe platform lets users monitor, limit, and stop AI agents that exceed their defined permissions.
- 01/10JEV-GW: an AI specialized in helping with decisionsThe JEV-GW presentation includes links to the event page and the project guide.
- 01/10Dot: the ChatGPT assistant that stays active in the cloudThe analysis gathers confirmed facts and issues raised in review videos about OpenAI’s agent for Pro and Business subscribers.
- 28/09agente-claude-codex 1.1.1 updates task records and checksThe update preserves history, checks only what ran, verifies sharing, and keeps memory up to date.
- 23/097 principles to manage AI agentsComparison of external material with INEMA’s methodology for managing AI agents.