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

Description and use

SkillHub workflow test verifying search, detail, and files without a live LLM succeeded after fixing Files tab timing, then replayed from cache with zero model calls.

Problem

Verify that SkillHub’s search, detail, and file pages work correctly without relying on a running model server.

Outcome

The test workflow succeeded after fixing the Files tab timing; replay with the model server stopped produced zero model calls and zero tokens, confirming that the workflow can run entirely from cache.

Workflow

• 1. Search for the "tdd" skill in SkillHub.
• 2. Open the skill in the "trend-tdd-en-hd" namespace.
• 3. Navigate to the Files tab and verify that SKILL.md appears.
• 4. Record the workflow with e2e 0.19.0 and @e2e-dev/web 0.14.0.
• 5. Stop the Spark server and run the recorded workflow in cache‑only mode.
• 6. Verify that all actions replayed with zero model calls and zero tokens.

Evidence

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

Source

The next complete workflow passed the search, detail route, active Files tab, and SKILL.md checks.

Source

Each agent.act reported zero model calls, the run used zero model tokens, and no physical model request receipt was written.

Source