Appendices

High-value references to support day-to-day use of this book and to accelerate adoption at scale.

Glossary

• AI-assisted development: Using AI tools (such as GitHub Copilot) to augment software delivery.
• Agent / agentic workflow: An automated sequence where the tool plans and executes steps with human oversight.
• Bounded context: A domain-driven design boundary within which a model is consistent.
• Characterisation test: A test that captures current behaviour to protect it during refactors.
• Chunking: Splitting work into small, reviewable changes that fit model limits and human review.
• CODEOWNERS: A repository mechanism to route reviews to specific teams or individuals.
• Instruction file: Repository-scoped guidance for Copilot (for example, .github/copilot-instructions.md).
• Prompt file: A reusable, versioned prompt for repeatable tasks (for example, .github/prompts/*.prompt.md).
• Strangler pattern: Incrementally replacing legacy capability by routing traffic to a new implementation.
• MTTR: Mean time to recovery; time taken to restore service after an incident.
• Change failure rate: Percentage of changes that cause incidents requiring remediation.

Quick reference checklists

Planning checklist

• Define scope, constraints, and acceptance criteria
• Choose chunking strategy (domain, layer, dependency, size, API, test)
• Establish documentation anchors (docs/architecture.md, standards) and link in instructions
• Ensure baseline tests exist for critical paths; add characterisation tests where needed
• Decide review gates, CODEOWNERS, and CI policy

Review and governance checklist

• Scope aligns to stated module/interface; small and focused
• Tests added/updated; CI green; coverage on critical paths acceptable
• Coding standards, naming, logging, and error handling followed
• Security/dependency scans clean; licences compliant
• Documentation updated (README/ADR/architecture note) as needed

Prompt library (templates)

Test expansion (module-focused)

• Goal: Increase unit tests for the specified module covering happy paths, error paths, and edge cases.
• Inputs: Source files in scope; existing tests; recent test output.
• Constraints: British English; do not change behaviour; follow naming and structure standards.
• Steps: Propose cases → generate tests → run → adjust as needed.

Refactor to standard logging

• Goal: Apply project logging pattern to the specified files.
• Inputs: Logging standard reference; files in scope.
• Constraints: Preserve behaviour; add structured error handling as per standards.
• Steps: Identify calls → replace/augment → add tests for error paths.

Markdown lint and language tidy

• Goal: Fix markdownlint issues and British English spelling/grammar in the specified file.
• Constraints: Do not change meaning; keep headings and lists well-formed.
• Steps: Lint → fix → re-run lint → summarise changes.

Tooling and commands

• Lint all Markdown (macOS/zsh):
• npx markdownlint '**/*.md'
• Build the static site (from repo root):
• python3 -m venv .venv && source .venv/bin/activate
• pip install -r requirements.txt
• ./build.sh

Further reading and resources

• GitHub Copilot documentation
• Copilot customisation (instructions and prompts)
• Domain-Driven Design (bounded contexts)
• Strangler Fig pattern (incremental modernisation)
• Contract testing with Pact

Key Takeaways

• Use the glossary and checklists to standardise language and execution across teams.
• Keep prompt templates short; point them to instruction files and high-signal docs.
• Treat documentation and tooling as part of the system; automate where possible.
• Validate outcomes with agreed metrics to sustain confidence and guide iteration.