Skip to content
Pere Villega
Go back

Agentic reflective practice

1 min read

Have you ever wondered why Claude Code sometimes completes a task with its allowed tools, but at other times reaches for a Bash command that requires permission?

It can buzz along with the built-in coding and editing tools, then suddenly try to pipe an obscure npx command into Python via tr. Is that a problem, and what can we do about it?

David Morgan-Brown explores those questions in Agentic Reflective Practice.


Share this post on:

Previous Post
Running Parallel Agents Without Collisions
Next Post
From One Agent to Many