Support multi-project workflow: applying context-infrastructure as a cross-project base while preserving target project's own `AGENTS.md` and skills

[dsdashun]

Context

I've been using context-infrastructure as a reference to build my own personal
context layer (SOUL, USER, axioms, skills, memory system). It works great as a
single workspace: I cd into it to start coding agent sessions, the root AGENTS.md loads, everything is
centralized.

But I've hit a workflow question that neither the README nor setup_guide.md
explicitly addresses, and I suspect other users will hit it too.

The problem

I want to use my personal context infrastructure while developing other,
pre-existing projects in my work (e.g. a web app, a Go service) that already have their own:

• AGENTS.md (architecture rules, commit policy, etc.)
• .agents/skills/ (project-specific skills)

The standard, correct way to develop such a project is to set the session's pwd
to the target project root and let the harness auto-load that project's
AGENTS.md and project-scoped resources. But the current design of
context-infrastructure assumes a single-workspace-as-pwd model (both the
README Quick Start and setup_guide.md imply you cd into the
context-infrastructure directory and work from there). This creates an apparent
either/or:

Option	pwd	Loads context-infra AGENTS.md?	Loads target project AGENTS.md?
A	context-infra directory	✅	❌
B	target project directory	❌	✅

On the surface it looks like you can only have one or the other.

Whar are the solutions for such scenario ?

[grapeot]

It's a good question. And the answer is, you don't need to do anything. Just put the project inside the context-infrastructure folder (anywhere will work), the harness will automatically load both AGENTs.md.
For example this is Codex's behavior: https://learn.chatgpt.com/docs/agent-configuration/agents-md Let me know if you hit other issues!