> Plan mode helped a bit. However, in Plan mode, Claude would write up a giant plan document and ask for feedback. It's hard to review a multi-page plan. Making matters worse, if you give it feedback, it would respond with a whole new version of the multi-page plan. That's not a productive way to plan out a project or feature.

It's sure baffling how Anthropic has kept Claude Code's plan mode so linear and inflexible. It may ask a couple of questions before writing it, but there's always going to be parts that need editing. Yet there's no good "Sounds good, but it needs these edits" option after it presents you the plan. It gives you (paraphrased) "1. Proceed with auto-edits 2. Proceed without auto-edits 3. Cancel the plan". Note that 3 doesn't even write the plan to the file at all, even if it's 95% fine. So your options are either A. Pick 1 or 2, immediately press escape to interrupt, then tell it to make edits or B. Pick 3, tell it to make edits, after which it has to write the entire plan from scratch again.

This is such bad UX that it really feels like either 1. Anthropic employees don't use Claude Code much - this seems incredibly unlikely or 2. It's intended to burn output tokens as it has to write a huge plan again.

IMO Superpowers isn't the ideal solution because it too lacks flexibility, but including the "plan sketch" stage is sure an improvement.

I've found [Plannotator]( https://github.com/backnotprop/plannotator) useful for this. It opens the plan on a minimal web page so I can highlight and add comments. This is then passed back to Claude code. Works well enough for now. But ideally this should be natively implemented in cc.

>It's sure baffling how Anthropic has kept Claude Code's plan mode so linear and inflexible

It's difficult to know what the appropriate process for a model would be without widespread deployment. I can see how they have to strike a fine balance between keeping up with what the feedback shows would be best and changing the way the user interacts with the system. Often it's easy to tell what would be better once something is deployed, but if people are productively using the currently deployed system you always have to weigh the advantage of a new method against the cost of people having to adapt. It is rare to make something universally better, and making things worse for users is bad.

There's also:

4. Tell claude what to do instead, which will update the plan base on what you say.

5. Add comments to the plan directly - similar to 4 - but you can comment on specific parts.

Note: I use the VSCode extension, not sure if it differs in terminal mode.

You can just tell it what to do. I have a cut and paste handy that lets the tool know to present all judgement calls to me a few at a time in logical groups to give feedback on. I go through that process and then it pulls the plan from that