How to drive openspec to "see that later" ?

[stevebeauge]

I've been using OpenSpec for a few months now, and I really like the workflow for daily work.

However, during explorations (/opsx:explore), I often run into new ideas or architectural decisions that don't really fit into the current change.

A real example: I'm building a local application that will scan the file system in its first release, then later support scanning remote sources through third-party APIs. During exploration, roadmap discussions naturally lead to architectural decisions that should influence the early implementation (for example, designing around multiple connectors). But other parts, like actually implementing those connectors, belong to future changes.

So far, the simplest approach I've found is to add a few cleanup/polish tasks in the task artifacts, explicitly asking to write docs/ADR or markdown spec files, then refer back to those files in later sessions.

It works, but it feels quite manual and happens often, especially on new applications where many things are still being discovered.

How do you usually handle this?

Is OpenSpec the right tool for long-term planning? Should I adapt OpenSpec to better fit this process, or combine it with other non-OpenSpec agents and use OpenSpec mainly for implementation steps?

[B2F]

Even if I didn't validate my exploration phase, I will build a proposal (if worth it). Then the task and "memory" is set in my "openspec list". Nothing prevents you to apply the explore phase on top of an existing proposal so that you can update it later.

You can even set the last task of the proposal artifact as a further @todo.

I tried Bmad, Superpowers, and from what I read speckit or spec kitty are neither as straight to the point as OpenSpec. If I cannot get the agent to understand me or to understand the code, I'll simply plugin an architect or senior dev skill from Bmad, while staying within the openspec framework.