Agentic Engineering Patterns: Linear Walkthroughs

• Source: https://simonwillison.net/guides/agentic-engineering-patterns/linear-walkthroughs/
• Author: Simon Willison
• Added: 2026-03-01
• Topic: Context engineering, agentic engineering, workflow reliability

Why this matters

Linear walkthroughs are a practical technique for reducing ambiguity in agent runs. By forcing a step-by-step, inspectable path through a task, they improve reproducibility, debugging speed, and team alignment.

Key takeaways

1. Break complex tasks into explicit sequential checkpoints.
2. Make each step produce a verifiable output artifact.
3. Keep transitions deterministic to reduce context drift.
4. Use walkthroughs as onboarding and review tools, not just execution aids.
5. Capture failure points per step to improve future runs.

Practical application notes

• Require a numbered execution plan before long agent tasks.
• Gate each step on a concrete assertion (file diff, test result, trace event).
• Persist walkthroughs as reusable runbooks for repeated workflows.
• Add timeout/rollback behavior at critical transitions.

Related notes

• agentic-engineering-patterns-code-is-cheap.md
• agent-reliability-chasm.md
• system-design-and-invariants-pattern.md