@hackproduct9: Most people think “keep the agent running” means one thing. It doesn’t. 🔁 There are two very different patterns: 🎯 /goal = condition-driven Run → evaluate → check condition → continue if needed. Best when the task has a clear finish line: migrations, refactors, backlogs, testable outcomes. ⏱️ /loop = time-driven Run → wait → run again. Best for: polling, recurring checks, monitoring, jobs with no natural end state. The mental model is simple: /goal asks: “Are we done yet?” /loop asks: “Is it time to run again?” And there’s a third mechanism worth knowing: Stop hooks — useful when you want logic to fire after every turn across sessions. One more subtle point: The evaluator doesn’t execute the work. It inspects the conversation and decides whether the goal condition has been satisfied. Condition or clock. Pick the trigger. 🧠 #ClaudeCode #AIAgents #AgenticAI #AIEngineering #DeveloperTools

1938

2026-08-20 05:02:04

2026-08-20 09:12:02

To see more videos from user @hackproduct9, please go to the Tikwm homepage.