Model calls first, model-off cache replay second: Spark-X2.5 tests SkillHub with e2e

说明与用途

使用 e2e 和 Spark-X2.5 对 SkillHub 进行测试后，修复了 Files 选项卡加载等待问题，证明可以通过 e2e 缓存完全重放工作流程，无需模型调用。

问题

Verify that SkillHub’s search, detail, and files workflow can be executed without live model calls by using native e2e cache replay.

结果

The workflow passed after model-assisted repair and then replayed with zero model calls, confirming that the native e2e cache can fully drive the SkillHub UI.

工作流

• Search for "tdd" in SkillHub
• Open the skill in namespace "trend-tdd-en-hd"
• Navigate to Files tab
• Verify SKILL.md appears

证据

The integration uses the documented StepExecutor extension. The custom executor observes the current page, supplies its real semantic controls and node IDs to Spark, and exposes one typing or clicking action for the requested step.

来源

The first full attempt produced a useful failure. Opening a skill changed the URL before the detail page was ready.

来源

I then stopped the owned Spark server and verified that loopback port 8098 refused connections. The same suite ran with the native cache in read-only mode. All four actions replayed. Each agent.act reported zero model calls, the run used zero model tokens, and no physical model request receipt was written.

来源