Adoption guide
Adoption should reduce context, not add another layer of duplicate rules.New repository
Start with the canonical templates from: github.com/datarelay-labs/engineering-system Base adoption requires:Existing repository with many AI rules
Do not blindly add another always-on rule. First classify the existing rules:Adoption-pending repositories
A valid Work Packet may exist before the repository has completed Engineering System adoption. WhenAGENTS.md or .engineering/* are missing during resume:
- record
ENGINEERING_SYSTEM_ADOPTION=ABSENT_OR_PENDING - continue under the canonical Engineering System default when the Work Packet is otherwise valid
- do not silently create, merge, or modify adoption files unless the active workstream explicitly authorizes adoption work
Install the Cursor user /resume command
On the machine running Cursor:Add ChatGPT routing, not the whole standard
ChatGPT Custom Instructions should stay short. They only need to say:- use the canonical Engineering System
- resolve the repository first
- use Work Packets for Cursor handoff and chat continuity
- read only minimal relevant context
- never claim unexecuted or different-HEAD evidence as PASS