Is it possible to integrate this skill into my own Python scripts?

Yes, agent.run provides a programmatic interface via the run_agent function, allowing for seamless integration into custom automation pipelines.

What is the difference between iterative and oneshot modes?

Iterative mode allows the agent to analyze results after each skill invocation and dynamically adjust its next steps, whereas oneshot mode plans all actions upfront in a single pass.

Can I run agent.run without an Anthropic API key?

Yes, the skill features a mock mode that generates intelligent simulated responses based on the agent type, allowing for local development and testing.

Where are the execution results and logs stored?

All executions are saved as structured JSON files in the agent_logs/ directory, including timestamps, prompts, skill results, and agent metadata.

How does the skill know which prompts to send to Claude?

The skill automatically generates system prompts by combining the agent's manifest data, capabilities, and workflow patterns into a Claude-friendly format.

Betty Agent Runner

Name: Betty Agent Runner
Author: epieczko

byepieczko

開発者ツール

Orchestrates the complete lifecycle of Betty agents by managing manifests, system prompts, and skill execution.

概要

The Betty Agent Runner is a sophisticated execution engine designed to power specialized AI agents within the Betty framework. It manages the entire agent workflow, from loading YAML manifests and generating Claude-optimized prompts to orchestrating API calls and executing planned skills. By supporting both iterative reasoning—where agents adapt based on feedback—and oneshot execution for predictable tasks, it provides a robust foundation for building autonomous development tools. The skill includes comprehensive JSON logging, diagnostic error handling, and a built-in mock mode for rapid development without API dependencies.

主な機能

Full agent lifecycle orchestration from manifest loading to execution
Support for iterative and oneshot reasoning modes for flexible task handling
0 GitHub stars
Automated translation of agent manifests into optimized system prompts
Comprehensive execution logging with structured JSON output for auditability
Built-in mock simulation mode for testing workflows without Claude API calls

ユースケース

Executing complex multi-step developer workflows like API design and specification validation
Automating repetitive coding tasks using specialized agent personas and personas
Debugging agent reasoning and skill selection using detailed execution logs

概要

主な機能

Full agent lifecycle orchestration from manifest loading to execution
Support for iterative and oneshot reasoning modes for flexible task handling
0 GitHub stars
Automated translation of agent manifests into optimized system prompts
Comprehensive execution logging with structured JSON output for auditability
Built-in mock simulation mode for testing workflows without Claude API calls

ユースケース

Executing complex multi-step developer workflows like API design and specification validation
Automating repetitive coding tasks using specialized agent personas and personas
Debugging agent reasoning and skill selection using detailed execution logs