🚀 xAPI-Scorable Interactive with Claude Code or OpenAI Codex external tools - SLS Interactive Response Questions

⚠️ Why web-based AI injection has limits for complex content

For complex simulations (EJS/EJSS, P5.js, custom JavaScript), reliable xAPI injection requires browser-based trial-and-error verification. This web tool analyzes your source code statically — it cannot run your simulation, observe what events actually fire, or confirm that window.storeState() receives correct data. The result is often generic tracking that silently sends score: 0.

✅ The correct approach — download and install an agentic AI tool:

• Claude Code — Download the installer (no command line needed to get started). Once installed, open your simulation folder in Claude Code. It can launch a Playwright browser, observe what events fire, generate tracking code, inject it, re-run, and iterate until storeState() sends real data.
• OpenAI Codex — Get Codex here. Same agentic workflow: run simulation → observe → generate → verify → fix — all in a local environment with full browser access.
• Visual Studio Code (Microsoft) — Download VS Code. With an agentic coding setup such as Copilot agent mode and local browser/terminal access, VS Code can support the same run → observe → edit → verify loop for xAPI injection.
• TRAE (ByteDance) — Download TRAE. Use its AI IDE or SOLO workflow to open the simulation project, modify the code, run checks, and iterate until real interaction data is captured.
• Google Antigravity — Download Antigravity. Google's agent-first development environment can orchestrate coding agents across a local project, making it suitable for trial-and-error xAPI integration work.

These tools run on your computer, can be given terminal and browser access, and can loop until the xAPI integration is verified working — something a web-based single API call cannot do.

Another reported workflow: plain ChatGPT GPT-5.5

Some users have reported success using plain ChatGPT with GPT-5.5 after using this webpage. Download the ZIP produced here, upload that ZIP file to ChatGPT, and ask ChatGPT to make the necessary xAPI edits inside the files.

Remember to ask ChatGPT to give you a downloadable ZIP file when it is done, then upload that ZIP to SLS. For complex simulations, still test the result carefully; if the tracking sends only score: 0, use one of the browser-capable agentic tools above.

💡 For standard content: Timeline Mode and Minimal Mode above work well without AI for quizzes with radio buttons, checkboxes, and common score patterns. Use those for most SLS content.