Codex CLI 0.128.0 adds /goal

2026-04-30 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, quick

Summary

OpenAI's Codex CLI coding agent has been updated to version 0.128.0, introducing a new "/goal" feature. This functionality allows users to define a specific objective, and the Codex agent will iteratively work towards its completion. The system continues processing until it self-evaluates that the goal is met or its pre-configured token budget is exhausted. This new feature, reminiscent of the "Ralph loop," is primarily driven by injected prompts such as "goals/continuation.md" and "goals/budget_limit.md" at the conclusion of each operational turn. The update was released on April 30th, 2026.

Key takeaway

For developers using coding agents for iterative tasks, the new "/goal" feature in Codex CLI 0.128.0 offers enhanced automation. You should experiment with defining explicit goals to leverage the agent's ability to self-evaluate and continue until completion or budget exhaustion, potentially streamlining development workflows for complex coding challenges.

Key insights

Codex CLI 0.128.0 introduces a "/goal" feature for autonomous, iterative task completion.

Principles

• Goal-driven autonomous agents
• Iterative self-correction loops

Method

The system injects "goals/continuation.md" and "goals/budget_limit.md" prompts at turn end to manage goal evaluation and token budget.

In practice

• Define clear, measurable goals
• Monitor token budget consumption

Topics

• Codex CLI
• OpenAI
• Goal-Oriented AI
• Ralph Loop
• Token Budget

Code references

• openai/codex

Best for: Machine Learning Engineer, AI Engineer, Software Engineer, Prompt Engineer

Related on AIssential

• Codex /goal: Beginner Guide to OpenAI’s New Autonomous Goal Feature — Codex CLI's new "/goal" command enables autonomous, self-evaluating agentic behavior for goal completion.
• Ralph-loop 2.0? The real autonomous coder is coming... — LLM-driven goal features enable AI agents to autonomously pursue complex, long-running tasks by continuously evaluating progress.
• How to write AI agent loops in Claude Code and Codex — AI agent loops enable autonomous, scheduled, or goal-driven prompting, extending beyond human-triggered interactions.
• When NOT to use /goal in Codex — Codex's /goal feature is best for complex, measurable objectives requiring multi-step investigation, not simple edits or vague tasks.
• The Codex feature that works while you sleep — Codex's "/goal" feature enables autonomous, long-running AI task execution by defining measurable outcomes and iterative self-correction.
• OpenAI Codex App: death of the VSCode fork, multitasking worktrees, Skills Automations — Agent-native coding interfaces and specialized open models are rapidly advancing, shifting developer workflows and evaluation paradigms.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.