Does this help with how the planer works?

[ScottChapman]

I see a number of skills, I think they are pretty generalized. But I frequently use the claude code planner to start any work, and it doesn't seem like it leverages anything that was generated?

[starkmarkus]

My understanding is that generated skills only help the planner if they are actually discoverable and selected/loaded by the agent during planning.

So if Claude Code’s planner starts from a very general prompt and does not explicitly pull in the generated Understand Anything skills, then I would not expect it to automatically use that context. In that case the skills may exist, but they are not really part of the planning context.

The practical thing I would check is whether the generated skills are installed in the location Claude Code expects, and whether they show up as available skills. Then I would test with an explicit instruction like:

Before planning, inspect and use the relevant generated skills for this repository.
Use them to understand the project structure, conventions, and domain concepts before proposing the plan.

If that changes the quality of the plan, then the issue is probably not the skill content itself, but that the planner is not automatically invoking it.

So I think the useful feature would be tighter planner integration: when starting a plan, automatically surface the most relevant generated skills or project understanding artifacts, instead of relying on the user to remind the agent to use them.

In short: yes, it can help planning, but only if the planner actually sees and uses the generated skills. If Claude Code planner ignores them by default, then this may need either explicit prompt guidance or better integration.

[ScottChapman]

I think I figured out an explicit way to do it, you can just use the "understand-anything" prompt directly in planning more. That actually works well!

[Lum1104]

Hi, I got a quick question:

Did you use the planer with /understand to initialize the graph, or with /understand-chat for the answering?

If it is the later one, I think the answer is yes.

[ScottChapman]

Yea, the latter.