[yuyixg]

Problem

A function that says back a sentence the model passed it (it sets StopCompletion and its result is the turn's reply) could only be mocked with a fixed resultContent. The turn's output was then the case author's own words, so every assertion on it, llmJudge included, tested the case rather than the agent.

Change

• A mock's resultContent may hold {{args.name}}, filled with that top-level argument of the call it answers (MockResultTemplate). A string argument goes in as it is, any other JSON value as its raw text, and a missing argument or non-object arguments as nothing. A result without placeholders is left untouched.
• The recorded ObservedToolCall.ResultContent is the filled text, as the agent saw it.

Why now

The RSP phone agent's identify_service now speaks the model's restatement followed by a fixed address check. Its agent-test mocks use {{args.restatement}} Before we get too far, ... with stopCompletion: true, so the restatement the model wrote is what llmJudge scores.

Tests

MockFunctionExecutorTests: placeholders filled from the call's arguments, the observed result matching what the agent saw, non-object or missing arguments filling nothing, and a JSON result with no placeholders left as it is. All AgentTesting unit tests pass (374).

🤖 Generated with Claude Code

[qodo-code-review]

Code Review by Qodo

Sorry, something went wrong

We weren't able to complete the code review on our side. Please try again manually by commenting /agentic_review on this PR.

[qodo-code-review]

PR Summary by Qodo

Render mock results from tool-call arguments

✨ Enhancement 🧪 Tests 🕐 10-20 Minutes

AI Description

• Allow mock results to include top-level arguments supplied by the agent's tool call.
• Record the rendered result so observed calls reflect what the agent received.
• Test substitution, missing or invalid arguments, and unchanged results without placeholders.

Diagram

graph TD
  Call["Tool call"] --> Match["Mock matcher"] --> Render["Result renderer"] --> Content["Agent message"]
  Render --> Record["Observed call"]
  Mock["Mock definition"] --> Match
  Mock --> Render
  Call --> Render

Loading

High-Level Assessment

Keep the dedicated, narrowly scoped renderer. A general template engine would add unnecessary capabilities and dependencies for top-level argument substitution; rendering in the executor also ensures the message and observation use the same result.

Files changed (4) +81 / -3

Enhancement (2) +43 / -2

MockFunctionExecutor.csApply rendered results to messages and observations +3/-2

Apply rendered results to messages and observations

• Renders a matched mock's result with the current call's arguments. Writes that same text to message.Content and the observed tool-call record.

src/Plugins/BotSharp.Plugin.AgentTesting/Runtime/MockFunctionExecutor.cs

MockResultTemplate.csAdd top-level argument substitution for mock results +40/-0

Add top-level argument substitution for mock results

• Replaces {{args.name}} placeholders with string values or raw JSON for other value types. Missing properties and non-object or invalid arguments produce empty substitutions; results without placeholders remain unchanged.

src/Plugins/BotSharp.Plugin.AgentTesting/Runtime/MockResultTemplate.cs

Tests (1) +34 / -0

MockFunctionExecutorTests.csCover rendered mock results and fallback behavior +34/-0

Cover rendered mock results and fallback behavior

• Tests substitution through the executor and confirms the observed result matches the message. Also covers missing, invalid, and non-object arguments and an unchanged result without placeholders.

tests/BotSharp.Core.UnitTests/AgentTesting/MockFunctionExecutorTests.cs

Documentation (1) +4 / -1

AgentTestCase.csDocument argument placeholders in mock results +4/-1

Document argument placeholders in mock results

• Documents that ResultContent can reference a tool call's argument using {{args.name}}.

src/Plugins/BotSharp.Plugin.AgentTesting/Models/AgentTestCase.cs