Chore: Build the Video.js Agent Evaluation Harness and Prompt Set

[mihar-22]

Context

A controlled comparison needs repeatable prompts, environments, captured tool use, and outcome scoring.

Outcome

Create the evaluation harness and prompt set for Video.js-only, Video.js with Mux, and negative Mux scenarios.

References

• https://app.notion.com/p/3e297a7f89d08179a055c710bfd3652d

[github-actions]

Planning recommendation: add this to the roadmap as post-launch discovery and assign P2; do not assign the current GA milestone. #2890 is the correct parent. Complete this harness before evaluating Mux recommendation/negative-case results in #2939 and before the MCP decision in #2941.