About Andy Agent Lab

What Andy Agent Lab covers, who it is for, and how the site is structured.

Andy Agent Lab is a technical site by Tamar Peretz focused on AI agent security, evidence-gated AI work, prompt engineering operating patterns, and reviewable AI system behavior.

Portrait of Tamar Peretz, author of Andy Agent Lab

What this site covers

The site combines operational assets and technical writing for people who need AI-assisted work and AI systems to be more reliable, inspectable, and easier to review.

The main content focus areas are:

  • AI agent security, trust boundaries, and control-plane failures
  • evidence-gated AI use, fact-checking, and verification workflows
  • prompt engineering patterns for controlled and reusable execution
  • architecture, memory, routing, and tool-execution review
  • model behavior and evaluation topics where fluent output can be misleading without evidence

Who this site is for

Andy Agent Lab is intended for readers who need a more controlled and reviewable way to work with AI, including:

  • practitioners using AI in daily work and needing stronger verification
  • engineers and reviewers working on agentic systems, tool use, orchestration, or architecture boundaries
  • researchers and technical writers producing evidence-gated outputs
  • teams that need reusable operating patterns instead of one-off prompts

How the site is structured

The site is organized into five content layers:

  • How-to — step-by-step procedures for common AI tasks and review flows
  • Prompt library — reusable prompt assets, workflow files, and components
  • Policies — operating rules that tighten evidence boundaries, output constraints, and fail-closed behavior
  • Articles — deeper technical analysis, explanatory writing, and topic-level arguments
  • Reference — models, diagrams, terminology, and supporting technical context

How this site is meant to be used

The intended usage order is:

  1. Start with How-to to choose the correct procedure for the task.
  2. Add a Policy when the run needs stricter evidence boundaries or output constraints.
  3. Use Prompt library when you need reusable prompt assets or workflow files.
  4. Use Articles and Reference when you need deeper technical context, models, or diagrams.

Operating orientation

This site is built around a consistent operating orientation:

  • prefer inspectable evidence over fluent but unsupported output
  • make review paths explicit before execution
  • separate procedures, rules, prompt assets, and deeper reference material
  • treat AI outputs as reviewable artifacts, not as authority by default

Author

Andy Agent Lab is written and maintained by Tamar Peretz.

Follow