skills/jinko-trial-viz/evals/evals.json

{
  "skill_name": "jinko-trial-viz",
  "evals": [
    {
      "id": 1,
      "prompt": "Create a Jinkō trial visualization for trial SID tr-demo with time-series plots for Drug and TumorVolume, scalar plots for AUC and Cmax, and show me how to sanity-check just those sections.",
      "expected_output": "Uses jinko-trial-viz, previews the mutation by default and requires explicit confirmation or --apply, resolves the trial with client.get_trial(trial_sid), creates a TrialVisualization via trial.create_empty_trial_visualization(...) (or the bundled trial_viz.py script), configures timeseries and scalars sections via typed subservices, then prints sanity diagnostics and fails on errors.",
      "files": [],
      "assertions": [
        "The output creates the TrialVisualization via trial.create_empty_trial_visualization(...) or the bundled trial_viz.py script, not a raw POST request.",
        "The output resolves the trial via client.get_trial(trial_sid) rather than requiring manual core/snapshot ids.",
        "The output configures both timeseries and scalars sections via their typed subservices.",
        "The output includes a sanity-check step scoped to timeseries and scalars.",
        "The output is dry-run by default and requires explicit confirmation or --apply before mutation.",
        "After mutation, the output prints diagnostics and treats sanity errors as failure."
      ]
    },
    {
      "id": 2,
      "prompt": "I already have a trial viz tv-demo. Update it to add a scatter plot comparing AUC vs Cmax across control and treated arms, then retrieve the resulting visualization JSON.",
      "expected_output": "Requires explicit confirmation or --apply, uses client.get_trial_visualization('tv-demo') then viz.scatter_plots.add_x_vs_y_plot('AUC', 'Cmax', arms=['control', 'treated']) (or the bundled trial_viz.py update command), prints post-update diagnostics and fails on errors, then fetches viz.content() afterward.",
      "files": [],
      "assertions": [
        "The output updates the existing TrialVisualization rather than creating an unrelated new trial.",
        "The output uses scatter_plots.add_x_vs_y_plot (or set_config with mode XvsY) rather than a raw PATCH payload.",
        "The output includes x variable AUC, y variable Cmax, and the control and treated arms.",
        "The output retrieves the visualization JSON after updating.",
        "The update is dry-run by default and runs post-update sanity before reporting success."
      ]
    },
    {
      "id": 3,
      "prompt": "Create a trial visualization in folder missing-analysis, but do not create that folder if it does not exist.",
      "expected_output": "Previews by default. On --apply, resolves the explicit folder by id or exact name and exits with an error if unresolved rather than falling back to the project root.",
      "files": [],
      "assertions": [
        "An explicitly requested unresolved folder is an error.",
        "The visualization is never created in the project root as a fallback."
      ]
    }
  ]
}