Claude Code's underused /goal command for autonomous agent loops

My notes

Summary

A short promotional clip claims Claude’s /goal command turns Claude into an autonomous loop: it takes an action, checks whether the goal condition is met, and keeps going if not. The creator says she ran it autonomously for ~5 hours to fix many issues, and frames it as the “next unlock” over prompting every step manually.

Key Insight

• Core mechanic (as described): action, then evaluate goal condition, then repeat until met. Removes the need to prompt each step by hand.
• Concrete data point: creator claims a single autonomous run lasted ~5 hours, fixing “so many issues” without intervention.
• Framing: positioned as a shift from step-by-step prompting to outcome-based delegation, you define the end state, the agent iterates toward it.
• Caveat: this is a 24-second promotional creator clip. The mechanic and 5-hour run are the creator’s claims, not independently verified here. Treat as a tactic to test, not a documented spec.