devsan
All work

Case study

Agent Playbook (experiments)

Reusable agent playbooks and automation patterns.

In progressSolo buildAI Agents2026
Agent Playbook (experiments) project cover: Reusable agent playbooks and automation patterns.

Summary

A sandbox for agent workflows: reusable playbooks, automation patterns, and LLM-execution flows. The lab where the public Agentic Playbook lessons start.

Problem

The public Agentic Playbook can only teach what I've actually validated. Before a pattern becomes a lesson, it has to survive real use on real projects. Experiments live somewhere; this is that sandbox.

Architecture

encode patternrun patternfeedback into next lessonSandipoperatorExperiments RepopatternsCoding AgentClaude + MCP
UserServiceAgent
Sandbox feeds the public playbook; only validated patterns get shipped.

Result

A working sandbox where agent patterns get tested before they make the public playbook. Patterns over one-off scripts; lessons only ship when the pattern survives use. Iterating, indexed against the public site, and finding its first real users via the lessons it informs.

Highlights

  • Patterns over one-off scripts
  • Tested before they make the public playbook

Have something similar?

Let's talk.

Get in touch