Design: Evaluate the Current Skill and Documentation Stack

[mihar-22]

Context

The shipped Video.js skill, Markdown pages, llms.txt, and bundled docs establish the baseline for any MCP decision.

Outcome

Measure task completion, correctness, retrieval behavior, time, cost, and citations using the current stack.

References

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

[github-actions]

Scope check: #2938 also evaluates the current stack, but it appears to own the controlled MCP comparison while this issue establishes the current-stack baseline. Please confirm that boundary before work begins so the two issues do not produce duplicate reports.

Planning recommendation: add this as post-launch discovery on Roadmap project #7; do not assign the current GA milestone. Recommend P2 and relate it to #2890. Treat #2894 (including open PRs #2883 and #2888) and #2925 as inputs, then use this baseline for #2938 and the decision in #2941. No clear epic linkage found.

[github-actions]

Merged #3081 adds production-only markdown_fetched telemetry for Markdown twins, llms indexes, and Accept-negotiated Markdown responses. Use it as an input to the retrieval-behavior baseline; task completion, correctness, time, cost, and citation evaluation remain in scope here.