For sysadmins, DevOps and platform engineers who want to call LLMs from Bash and PowerShell, use terminal coding agents safely, and put AI steps into CI/CD without handing over the keys.
The curriculum is being prepared and reviewed. Enrollment is not open yet.
About this course
Most operational automation still lives in shell scripts, PowerShell modules and pipeline YAML. This course shows how to bring language models into that world without making it less safe or less predictable.
You start with the mechanics: calling vendor APIs from Bash with curl and jq and from PowerShell 7 with Invoke-RestMethod, requesting structured output and validating it before any script acts on it, and handling rate limits, timeouts and API keys correctly. You then work with terminal-native coding agents (Claude Code, GitHub Copilot CLI, Gemini CLI, Codex CLI, Kiro CLI) in interactive and non-interactive modes, including permission and allowlist settings. The second half moves into pipelines: AI-assisted log triage and change review in GitHub Actions and GitLab CI, generating and refactoring scripts with lint and test gates (ShellCheck, PSScriptAnalyzer, bats-core, Pester), and a guardrail model for any automation that turns model output into commands. Tool names and modes are current as of October 2026; the patterns are designed to outlive specific CLIs.
What you will be able to do
Call an LLM API from Bash and from PowerShell 7 with structured JSON output, schema checks, retries with backoff and keys held outside the script
Run a terminal coding agent non-interactively in a script or pipeline with a restricted tool allowlist and an auditable log
Build an AI-assisted log and incident triage step that summarizes evidence without executing anything on its own
Add an AI review or generation job to a GitHub Actions or GitLab CI pipeline with least-privilege tokens and human approval gates
Generate or modernize shell and PowerShell scripts and prove them with ShellCheck, PSScriptAnalyzer, bats-core and Pester before merge
Design a guardrail policy (dry-run, allowlists, sandboxing, approval) for any automation that turns model output into commands
Syllabus
6 modules · 18 lessons · about 12 hours
1
Module 1: Calling Models from the Shell
The plumbing: HTTP calls, JSON in and out, and secrets handled like any other production credential.4 lessons
Bash, curl and jq: A Robust LLM Call in Forty Lines
PowerShell 7 and Invoke-RestMethod: Objects In, Objects Out
Structured Output in Scripts: Schemas, Validation and Failing Closed
API Keys, Rate Limits and Retries in Unattended Jobs
2
Module 2: Terminal Coding Agents
What the agentic CLIs do, how they differ, and how to run them headless without giving them the whole machine.3 lessons
The 2026 Terminal Agent Landscape: Claude Code, Copilot CLI, Gemini CLI, Codex CLI and Kiro CLI
Permissions, Allowlists and Sandboxes for Agents in a Shell
Headless Mode: Scripting an Agent Inside a Larger Job
3
Module 3: AI for Operations Work
Day-to-day use: logs, runbooks and incident evidence, with the model as analyst rather than operator.3 lessons
Log and Journal Triage: Summarizing Without Leaking
Turning Runbooks into Assisted Procedures
Natural Language to Command: Explain First, Execute Never by Default
4
Module 4: AI Steps in CI/CD
Putting model calls into pipelines with the same discipline as any other build step.3 lessons
AI Review Jobs in GitHub Actions and GitLab CI
Least-Privilege Tokens and Prompt Injection Through Pull Requests
Cost, Caching and Determinism in Pipeline AI Steps
5
Module 5: Generating and Modernizing Scripts
Using AI to write and refactor automation, and proving the result with linters and tests.3 lessons
Generating Scripts You Can Trust: ShellCheck and PSScriptAnalyzer as the First Reviewer
Test Harnesses for AI-Written Automation: bats-core and Pester
Porting Legacy Bash and Batch to PowerShell 7 or Python with an Assistant
6
Module 6: Guardrails and Operating Model
A policy you can defend for automation that acts on model output.2 lessons
The Guardrail Ladder: Suggest, Dry-Run, Approve, Execute
Audit Trails, Change Records and Rolling Back AI-Initiated Changes
Exam and certificate
The exam unlocks once you have completed every lesson. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Anthropic (Claude Code), GitHub (Copilot CLI, Actions), Google (Gemini CLI), OpenAI (Codex CLI), Amazon (Kiro CLI), Microsoft (PowerShell, AI Shell), or GitLab (Duo). Product and company names are trademarks of their respective owners.
A free, one-hour, hands-on start built on one realistic job: sorting a property manager's 1,150 monthly maintenance requests. You get an API key the safe way, make real calls with curl and an official Python SDK, run them over forty work orders, read stop reasons, usage and errors, estimate the monthly cost, and decide whether the script is ready. Exercises run on your own provider account for well under a dollar.
A code-first course on integrating LLMs into Python the way production code needs it: uv projects and secrets, the OpenAI, Anthropic and Google GenAI SDKs side by side, async concurrency and streaming, Pydantic structured outputs and tools, retries, local models, embeddings and RAG, FastAPI and pandas integration, testing and evals with pytest, observability, cost tracking and deployment.
Call OpenAI, Anthropic, Gemini and local Ollama models from Perl with real CPAN modules, add streaming, concurrency and validated JSON output to text pipelines, and use AI coding assistants to modernize legacy Perl behind characterization tests.