Fix out-of-range @ref result references in program interpreter

[Tal Zaccai (TalZaccai)]

Problem

Result references ({ "@ref": n }) were validated only with typeof index === "number" && index < length. A negative index like -1 passed the check, and results[-1] silently resolved to undefined — substituting undefined where a validated value from a preceding step was expected. createModuleTextFromProgram had the same gap. evaluateJsonProgram is a public export usable without validation, so this is reachable in practice.

Fix

• Add Number.isInteger(index) && index >= 0 && index < … bounds checks in both evaluateJsonProgram and createModuleTextFromProgram.
• In the interpreter, throw Invalid result reference on a bad index instead of falling through to an implicit undefined — otherwise it would still silently yield undefined.

Tests

• Added typescript/test/program.test.ts covering valid, negative, non-integer, and out-of-upper-bound references for both functions, and wired it into the test script.
• Full suite passes (87/87).

[Copilot]

Pull request overview

This PR tightens validation and runtime handling of result references ({ "@ref": n }) in the TypeChat program interpreter and module generator so invalid indices (negative/non-integer/out-of-range) no longer silently resolve to undefined.

Changes:

• Strengthen @ref bounds checks in evaluateJsonProgram and createModuleTextFromProgram to require integer, non-negative indices within bounds.
• Make the interpreter throw an explicit Invalid result reference error on invalid indices.
• Add a new TypeScript test file for @ref behavior and wire it into the typescript package test script.

Show a summary per file

File	Description
typescript/src/ts/program.ts	Adds stricter @ref validation and throws on invalid references during interpretation.
typescript/test/program.test.ts	Introduces tests covering valid and invalid result references for both APIs.
typescript/package.json	Updates the test script to execute the new compiled test.

Review details

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 3/3 changed files
• Comments generated: 3
• Review effort level: Low