Tool composition patterns — CCA-F Exam Prep

L2.08✨ 0

REAL STORY A senior engineer at a whiteboard drawing a pipeline: four boxes connected by arrows labeled 'search -> extract -> validate -> store.' She's pointing at the arrow between extract and validate, explaining something to two junior devs. The whiteboard has sticky notes: 'What if validate fails?' and 'Don't skip steps.'

A senior engineer at a whiteboard drawing a pipeline: four boxes connected by arrows labeled 'search -> extract -> validate -> store.' She's pointing at the arrow between extract and validate, explaining something to two junior devs. The whiteboard has sticky notes: 'What if validate fails?' and 'Don't skip steps.'

A lead engineer built a four-tool pipeline. It processed 10,000 records in 3 minutes. Then one tool failed and corrupted the entire dataset.

Her team built a data enrichment agent. It searched a vendor API for company data, extracted relevant fields, validated the format, and stored results in the database. Four tools, chained together. Each tool's output was the next tool's input.

It worked beautifully for two weeks. Then the vendor API changed a field name. The extract tool returned mangled data. The validate tool was checking types, not field names. The store tool saved 4,000 corrupted records to production.