AI Agent Playbook

Python utilities and examples for designing safer AI-agent workflows, risk checks, and human-in-the-loop approval boundaries.

The goal is not to build a large agent framework. The goal is to make common automation decisions explicit: what pattern fits the task, what risk level applies, and when a human should approve the next action.

Features

• Define reusable agent workflow patterns
• Classify task risk before automation
• Generate lightweight workflow plans
• Keep human approval for high-risk actions
• Provide small, testable Python examples

Example use cases

• Planning an AI-assisted research task
• Deciding whether an automation can run directly
• Documenting approval gates for multi-step workflows
• Building small internal playbooks for agent operations

Project structure

ai_agent_playbook/
  cli.py       # CLI entry point
  patterns.py  # reusable workflow patterns
  risk.py      # risk levels and approval gates
tests/
  test_risk.py

Run locally

python3 -m ai_agent_playbook.cli "research a company" --risk medium
python3 -m ai_agent_playbook.cli "send outreach emails" --risk high

Run tests

python3 -m unittest discover -s tests -p 'test*.py'

Example output

Pattern: research-brief
Control: Dry-run and verify before: research a company

Design principles

• Prefer small workflows that can be inspected
• Keep evidence and outputs traceable
• Use dry-runs before irreversible actions
• Require explicit approval for high-risk work
• Treat agents as workflow helpers, not magic workers

License

MIT