Course Objective & Scope
The same frontier LLM can flail endlessly in one repository while completing complex multi-file tasks cleanly in another; the difference rarely lies in the model weights, but in the execution harness around the agent. This course teaches how to engineer the tools, verifiers and environment that make coding agents reliable at enterprise scale.
Participants build fast agent-friendly verification commands, low-noise compiler/test output formatters, AST/LSP-backed search and edit tools, deterministic architectural linters and hooks, state checkpoint/rollback mechanisms, and an internal SWE-bench-style evaluation suite on a real codebase.
Who Should Attend?
- Platform, Developer Experience (DevEx) and AI infrastructure engineers
- Senior/staff software engineers and software architects
- Teams building or customising internal coding agents and SDLC automations
- Prerequisite: strong software engineering background, CLI scripting and CI/CD infrastructure experience
Key Takeaways
- Diagnose whether agent failures stem from model limits or missing harness affordances
- Write tool wrappers that return compact, actionable error diagnostics without flooding context
- Set up sub-second targeted test, typecheck and architectural boundary verifiers
- Enforce deterministic guardrail hooks that block unsafe edits and forbidden commands
- Build an internal benchmark from historical PRs to measure harness improvements objectively
Curriculum
Harness Engineering Concepts and the Agent-Computer Interface (ACI)
- Why harness design dominates raw model swaps; lessons from SWE-bench and production
- ACI design principles: observation space, action ergonomics and high-signal diagnostics
Optimising Search, Navigation and Editing Tools
- Exposing ripgrep, AST (tree-sitter) and LSP symbol queries cleanly to the agent
- Edit primitives that prevent line-drift and malformed diff failures on large files
Fast, Low-Noise Verification Harnesses
- Output wrappers that strip multi-thousand-line logs down to root errors and file:line pointers
- Incremental test and typecheck runners that verify touched code in seconds
Deterministic Guardrails, Hooks and State Recovery
- Pre-tool and post-edit hooks blocking layer violations and test tampering
- Automated Git snapshots, workspace isolation and clean state rollback
Internal Agent Benchmarks and Evaluation Suites
- Turning historical bug-fix and feature PRs into reproducible task containers
- A/B testing harness changes using pass@1, token cost, step count and human-intervention metrics
Other Courses in This Category
-
Advanced training on designing closed-loop autonomous coding cycles (generate → compile → test → observe logs/traces → repair) with convergence criteria, budget caps, anti-reward-hacking rules and nested loop architectures.
- Live Course
- 5 modules
-
Combining static analysis metrics with LLMs to understand unfamiliar codebases, detect technical debt and code smells, execute safe refactoring backed by characterization tests, and optimise performance.
- Live Course
- 5 modules
-
Developer training on generating unit, integration, contract and property-based tests with AI, using mutation testing to verify that generated test suites actually catch defects rather than mirroring implementation bugs.
- Live Course
- 5 modules
-
Solving complex software bugs through hypothesis-driven debugging with logs, stack traces, OpenTelemetry traces, git bisect and minimal reproducible examples (MREs) instead of blind trial-and-error prompting.
- Live Course
- 5 modules
-
Building living technical documentation directly from codebases: Architecture Decision Records (ADRs), C4/Mermaid diagrams, OpenAPI/SDK references, onboarding guides and automated doc-drift checks in CI/CD.
- Live Course
- 5 modules
-
Advanced training on repository-scale context engineering for monorepos and large codebases: AST indexing, hybrid search, dependency graphs, hierarchical AGENTS.md/CLAUDE.md files and context-window budgeting.
- Live Course
- 5 modules
Get in touch
