Descriptions:
Web Dev Cody walks through the new /goal command available in both Claude Code and OpenAI Codex CLI, a feature that implements an autonomous loop enabling the agent to keep iterating toward a defined objective without requiring manual prompting or approval at each step. Unlike standard sessions that stall and wait for user input, goal mode runs continuously until the task is marked complete — or until explicitly paused with commands like /goal pause, /goal resume, or /goal clear.
The video demonstrates two concrete use cases from the creator’s own Mission Control project: using /goal to get a GitHub Actions workflow fully passing for an Electron app signing pipeline (a notoriously tedious back-and-forth process), and using it to address every item surfaced by a codebase security and performance audit without babysitting each fix. The creator notes that Codex’s implementation is still experimental, requiring manual activation through /experimental settings, while Claude Code’s version appears more natively integrated.
A key workflow tip covered is combining goal mode with custom “skills” — reusable prompt templates that enforce production-quality coding standards — allowing developers to kick off long autonomous sessions, walk away, and review a complete diff only when finished. The video is short and practitioner-focused, making it a useful quick-start reference for developers already working in Claude Code or Codex who want to reduce manual interruptions on long refactoring or CI tasks.
📺 Source: Web Dev Cody · Published May 13, 2026
🏷️ Format: Tutorial Demo
