Exercise 08: The Agentic Pipeline — The Agentic Crew Hands-On Guide

Overview

You've been the one typing prompts. In every chapter so far, you started the conversation. You were the driver. The agent was the engine — powerful, fast, tireless — but it only moved when you turned the key. This chapter is about removing yourself from the loop.

You'll build an automated multi-step pipeline: unstructured feedback goes in, AI agents parse, categorise, and prioritise it, GitHub issues come out, and a summary report is generated — all orchestrated by shell scripts and scheduled with cron.

This is 'multi-agent' orchestration. Two agents, two system prompts, two jobs. The intake agent doesn't know about priorities. The triage agent doesn't do parsing. Each one is specialised, and the shell script connects them. You've just built what framework vendors charge you a monthly subscription for — with two bash scripts and a pipe.
— from the exercise

What you'll build

By the end of this exercise

✓ An Intake Agent that parses unstructured feedback into structured JSON

✓ A Triage Agent that categorises items (bug/feature/praise) and assigns priorities

✓ An Issue Creator that automatically files GitHub issues with proper labels

✓ A Summary Agent that generates weekly Markdown reports with trends

✓ A pipeline script that orchestrates all agents in sequence

✓ A cron schedule (or GitHub Actions workflow) that runs it all automatically

Key concepts

Terms you'll learn

Term What it means

Pipeline A chain of steps where each step's output feeds into the next step's input

--print flag Makes the CLI output go to stdout — no interactive conversation, composable with pipes

System prompt Instructions that define an agent's role and behaviour for a specific task

Multi-agent Multiple specialised agents, each with one job, connected by scripts

cron A Linux scheduler that runs commands at specified times (e.g., every Monday at 8 a.m.)

Triage Sorting incoming items by type (bug, feature, praise) and priority (high, medium, low)

What's covered

Sections in this exercise

01 What You'll Build

02 What You'll Need

03 The Boring Work

04 Building Block One: The Intake Agent

05 Building Block Two: The Triage Agent

06 Wiring the Pipe

07 Connecting to GitHub

08 The Summary Report

09 Set It and Forget It

What just happened

You built an automated pipeline from nothing but shell scripts and AI agents. The intake agent parses, the triage agent categorises, issues get filed, reports get written — all running on a schedule while you sleep. You designed the system. The agents execute it.