What the course covers
- Module 1: Scope it · about 75 minutes
You keep: The starter kit, a one-page agent design template, and the worked design for the invoice desk - Module 2: Tools · about 80 minutes
You keep: Tool design checklist, and the worked solution to the lab (narrow tools that keep account numbers from the model) - Module 3: The loop · about 80 minutes
You keep: Loop design worksheet (limits, approvals, failures, hand-over), and the worked solution to the lab (a repeat guard and a hand-over) - Module 4: Memory and data · about 85 minutes
You keep: Agent data map template, and the worked solution to the lab (policy search with sources, and trace retention) - Module 5: Test it · about 90 minutes
You keep: Evaluation plan template, and the worked solution to the lab (a noisy model, a process check and an evaluation gate) - Module 6: Attack it · about 85 minutes
You keep: Red-team playbook mapped to OWASP, and the worked solution to the lab (scripted attacks with a hijacked model, an independent harm judge, scope and output checks) - Module 7: Watch it · about 80 minutes
You keep: Monitoring plan (measures, alerts, review sample, drift, stop switch, incidents), and the worked solution to the lab (measures from traces, alerts and a budget stop) - Module 8: Ship it · about 80 minutes
You keep: Go-live checklist, rollout plan and runbook template, and the worked solution to the lab (stages, shadow mode, cost model, rollback) - Capstone: a production-ready agent, with its evidence · about 10 hours
Then the final test: 30 questions, 27 to pass.
Two ways to take it
- Online, at your own pace. Saolix gives each client organisation one access code for its learners. Enter it below to open the course in this browser. Your certificate states that the capstone was self-assessed.
- In a classroom. For 10 or more learners from one organisation, a Saolix trainer teaches the course, assesses each capstone and gives feedback. Ask for dates and pricing.
Open the course
The code is checked in this browser, which then remembers it so you need not type it again. Nothing is sent to Saolix. No code? Ask your organisation's training lead, or contact us.
30 questions
Answer all 30 questions, then check your answers. You need 27 out of 30 to pass and receive a certificate. Each attempt draws a fresh set of 30 from a bank of 100. If you score less, you will see which questions you got wrong and the module to revisit for each, then retake the final test. Your answers, progress and times are kept only in this browser.
Your answers
Your certificate of completion
Enter your name as you want it to appear, then save the certificate as a PDF. In the print window, choose Save as PDF. A certificate is issued once per completion of the course.
Rate this course
How would you rate this course? Your rating is anonymous — just a number, with no name and no account.
Saolix does not record who takes this course, so it cannot verify these certificates. The certificate confirms completion of a self-paced course and is not an accredited qualification.
Standards and frameworks covered
The published frameworks this course teaches toward. Saolix is not affiliated with, endorsed by or accredited by any of these bodies; naming a standard is a citation, not a claim about its author.
Sources
The official documents this course relies on. Laws and guidance change, so check the current version.
- Data Protection Act, 2019 (sections 25 and 48 to 50) · Kenya Law
- OpenAI compatibility (the chat-completions endpoint and tool support) · Ollama
- Python downloads and documentation · Python Software Foundation
- Building effective agents (workflows and agents; start with the simplest solution) · Anthropic
- OWASP Top 10 for Agentic Applications for 2026 (the principle of least agency) · OWASP GenAI Security Project
- Mandate fraud (changed supplier bank details) · National Business Crime Centre (UK police)
- OWASP Non-Human Identities Top 10 · OWASP Foundation
- JSON Schema reference (types, enum, pattern, minimum and maximum) · JSON Schema
- Python documentation, unittest (writing the lab's tests) · Python Software Foundation
- Data Protection (General) Regulations, 2021 (regulations 19, 22, 26, 37, 40, 46 and 49) · Kenya Law
- Text Embeddings Reveal (Almost) As Much As Text (Morris and others, EMNLP 2023) · arXiv
- τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains (Yao and others, Sierra, 2024), the source of pass^k · arXiv
- Adding Error Bars to Evals: A Statistical Approach to Language Model Evaluations (Miller, 2024) · arXiv
- Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena (Zheng and others, 2023), on the biases of model graders · arXiv
- Demystifying evals for AI agents (2026) · Anthropic
- Glossary (temperature: even at 0.0, results are not fully deterministic) · Anthropic
- Python documentation, random (seeded random numbers for repeatable tests) · Python Software Foundation
- Python documentation, unittest.mock (replacing a function during a test) · Python Software Foundation
- OWASP Top 10 for Large Language Model Applications 2025 · OWASP GenAI Security Project
- What is OpenTelemetry? (an open-source, vendor-neutral toolkit for traces, metrics and logs) · OpenTelemetry (Cloud Native Computing Foundation)
- Model IDs and versioning (pinned versions, aliases and serving changes) · Anthropic
- Gemini API models (stable versions and "latest" names) · Google
- Python documentation, hashlib (stable hashes) and copy (deep copies) · Python Software Foundation