Feature request: explicit skill-only mode for complex environments

First, GStack is great. I like the structure, the depth of the reviews, and the idea of carrying a project from objective discovery through implementation.

My environment is unusually large and complex, though, and I believe the workflow lost track of important context somewhere between initialization, objective questioning, design, review, and implementation. With all of those stages and their associated context active, environment-specific constraints and earlier decisions were sometimes missed or misdiagnosed.

I also found that I was unable to use several other Codex and Claude skills that I normally rely on. In practice, GStack became the primary orchestration layer instead of one composable tool within the host environment.

Is there a recommended way to run GStack in a strictly explicit, skill-only mode, where individual capabilities can be invoked without GStack becoming the front-running workflow or carrying unrelated context into later stages?

I am experimenting with a small fork/derivative of your design for users who prefer explicitly invoked skills inside an existing Codex or Claude workflow. This is not a criticism of GStack's opinionated approach or a request to remove what you built; it is a different integration point for larger environments. I will keep you posted as that develops.

Start by reviewing how GStack carries context across initialization, objective questioning, design, review, and implementation. Compare that flow with the requested explicitly invoked, skill-only use inside Codex or Claude; done would mean defining a clear integration behavior that avoids becoming the front-running workflow or carrying unrelated context forward.

Written by the indexing model from the issue text.