Increase toolCallBudget of some hosted-copilot-sdk invocation tests

[JasonYeMSFT (JasonYeMSFT)]

Description

The modified tests expect the agent to view the codebase to find out it needs to invoke the azure-hosted-copilot-sdk skill. Reading files in the workspace uses tool calls. The default 3 tool call budget is not enough for the agent to read the files and invoke the skill. Extending it to 10.

Checklist

• Tests pass locally (cd tests && npm test)
• If modifying skill descriptions: verified routing correctness with integration tests (npm run test:skills:integration -- <skill>)
• If modifying skill USE FOR / DO NOT USE FOR / PREFER OVER clauses: confirmed no routing regressions for competing skills
• Version bumped in skill frontmatter (if skill files changed)

Related Issues

[Copilot]

Pull request overview

This PR increases the allowed “early tool calls” window for the azure-hosted-copilot-sdk integration tests so the agent can scan the workspace (via tool calls) before being required to invoke the expected skill.

Changes:

• Extend shouldEarlyTerminateForSkillInvocation to accept an optional toolCallBudget override.
• Update the azure-hosted-copilot-sdk integration tests to use a higher budget (10) for scenarios that rely on codebase scanning.

Show a summary per file

File	Description
tests/utils/evaluate.ts	Adds an optional toolCallBudget parameter to control the early-termination threshold per test.
tests/azure-hosted-copilot-sdk/integration.test.ts	Increases the early-termination tool call threshold to 10 for the “existing SDK app” scenarios.

Copilot's findings

Comments suppressed due to low confidence (1)

tests/azure-hosted-copilot-sdk/integration.test.ts:116

• Same as above: consider using a named constant instead of repeating the 10 tool-call budget literal, so the value can be tuned in one place.

        const rate = await measureInvocationRate(agent, SKILL_NAME, {
          setup: setupCopilotSdkApp,
          prompt: "Add a new feature to this app that summarizes pull requests",
          shouldEarlyTerminate: (agentMetadata) => shouldEarlyTerminateForSkillInvocation(agentMetadata, SKILL_NAME, 10),
        }, "existing-sdk-modify", RUNS_PER_PROMPT);

• Files reviewed: 2/2 changed files
• Comments generated: 2

[Jon Gallant (jongio)]

Clean test tuning. The optional toolCallBudget keeps the default behavior for every other callsite, and raising to 10 matches the two setupCopilotSdkApp scenarios where the agent has to scan the workspace before it can reasonably invoke the skill. The other three tests in the same file don't need the bump, so the targeted split is right.

Agree with your call on keeping 10 inline until there's a third usage, and on skipping input validation for a 2-callsite internal helper.