Engineering Playbook
CONTEXT
Give engineers and AI the right information before asking them to solve the problem.
High-performing engineering starts with shared understanding. Context reduces rework, ambiguity and fragile AI-assisted output.
What breaks without it
Without shared context, teams and AI tools optimise locally and create rework, inconsistency and fragile delivery.
The deliverable
The Context Pack
Context stops being advice and becomes infrastructure.
A context pack is the machine-readable environment surrounding a repository or team. It exists so that a capable engineer — or an agent — can start work and make a good decision without waiting for a meeting.
Its real purpose is to paint a picture of what good looks like. Everyone touching the system should be working to the same expectation rather than quietly inventing their own — and that includes the contributors who are not human.
The test of a context pack is simple: hand it to someone who has never seen the system and see whether they reach the same conclusion your senior engineers would.
- Product intent and domain glossary
- Architecture map and decision records
- Coding standards and approved libraries
- Security constraints and data classification
- Definition of Ready and Definition of Done
- Architectural boundaries and ownership
- Test commands and environment setup
- Repository instructions and agent skills
- Tool and integration permissions
- Production constraints
Practices
- Treat context as a maintained asset with an owner, not documentation debt
- Keep instructions close to the code they govern, not in a distant wiki
- Prefer a short decision record over a long forgotten page
- Make non-goals as discoverable as goals