Agentic Engineering
The discipline of shipping production software with coding agents. The agent loop, delegation, spec-driven development, agent-assisted testing, review and plan patterns, context engineering, tool servers, legacy code, debugging, and fleet governance.
FundamentalsTopics 1–10
- ·From vibe coding to agentic engineering
- ·Where speed without control breaks
- ·Author to conductor
- ·The agent loop
- ·Tools are messages, not hooks
- ·Tool output shape
- ·Build the loop yourself
- ·When the first agent misbehaves
- ·The delegation framework
- ·Mapping delegation to your team
Start Fundamentals →
IntermediateTopics 11–20
- ·Specs with teeth
- ·The five phases
- ·The constitution
- ·Specify, and its exit test
- ·Clarify, the cheap phase
- ·Incremental or batch
- ·Acceptance criteria as tests
- ·Negative paths first
- ·Vacuous tests and the review checklist
- ·The mutation check and the CI gate
Start Intermediate →
AdvancedTopics 21–30
- ·Two axes, one harness
- ·Plan mode as an approval gate
- ·Where the split helps and hurts
- ·Fresh-context review
- ·Reviewer prompts and triage
- ·Skills and slash commands
- ·Sub-agents and least privilege
- ·Parallel sessions and model tiers
- ·Context engineering
- ·Taking control of your MCP surface
Start Advanced →
ProductionTooling & Governance — Topics 31–42
- ·No fear and no memory
- ·Characterisation tests
- ·The safe refactoring loop
- ·The cost of shape
- ·Promote, discourage, forbid
- ·The deliberate-violation test
- ·Microtooling
- ·The alert-to-fix pipeline
- ·Hypothesis fan-out, red first
- ·Governing the fleet
- ·The leadership conversation
- ·Running models locally
Start Production →