Overview

Defines how to interact with Agentica's async REPL harness during testing. It enforces awaiting futures and delivering results in a single code block to keep execution deterministic.

How This Skill Works

In practice, any function that returns a Future must be awaited (e.g., content = await view_file(path)). Each response should consist of one code block; multiple blocks will trigger only the first to execute. This protocol keeps test runs consistent and predictable when using Agentica's async REPL harness.

When to Use It

• When testing tools that return futures (e.g., view_file, ask_memory) in the async REPL harness
• When you need deterministic execution where only a single code block runs per response
• During integration tests for Agentica prompts and memory queries
• When benchmarking or validating end-to-end async flows
• When documenting expected REPL behavior for teammates

Quick Start

1. Step 1: Identify futures-returning calls (e.g., view_file, ask_memory) and use await
2. Step 2: Write a single coherent code block that computes and returns the final value
3. Step 3: Run in the async REPL harness and verify the output matches expectations

Best Practices

• Always use await for any Future-returning tool
• Return results in a single code block; avoid extra blocks
• Keep the code block self-contained and readable
• Test both successful and edge-case inputs in one block
• Document any quirks of the REPL in comments

Example Use Cases

• content = await view_file(path)
• answer = await ask_memory("What's the status?")
• content = await view_file(path) return any(c.isdigit() for c in content)
• # BAD: Split blocks (second block never runs) content = await view_file(path) return len(content) > 0
• content = await view_file(path) return content.upper()

Frequently Asked Questions