idkmesh

External Project Adoption Guide — 2026-09-22

Project-owner request

Clarify how IDKMesh should be used to develop a separate application in a collaborative environment. The requested path starts from creating a directory and GitHub repository, then connects LLMs, agents, humans, and tools; decomposes work into small chunks; routes work to appropriate participants; verifies results; and explains best practices in a machine- and agent-friendly way.

Repository interpretation

The repository already contained the core pieces needed for this explanation:

The missing element was a clear external-project onboarding path that connected these pieces from an empty project through day-to-day development.

Changes made

Added docs/PROJECT_ADOPTION_GUIDE.md with:

Updated the repository front doors so the guide is discoverable from:

Key design conclusion

The clearest adoption model is:

GitHub project state
  -> bounded Work Unit
  -> capability/risk routing
  -> human or agent candidate
  -> exact-revision evidence
  -> independent verification
  -> protected integration decision
  -> observed outcome
  -> next Work Unit

Provider/model selection comes after project policy, capability requirements, and risk classification. Larger models do not gain additional authority.

Current capability boundary

The guide intentionally does not claim a finished external-project bootstrap or universal live model dispatcher. ProjectManifest/DomainPack validation and adapter infrastructure exist, but a polished idkmesh init + multi-provider Verified Swarm Runner remains future product work.

Community impact

The change reduces the amount of architecture knowledge required for a new maintainer or contributor to answer “how do I use IDKMesh on my own project?” and gives humans and agents a shared operational vocabulary without requiring private project context.