| name | visualize |
| description | Build results dashboard, comparison view, or project overview. Triggers on "visualize", "可视化", "show results", "dashboard", "compare experiments", "chart my results", "project overview".
|
| disable-model-invocation | true |
Visualize
Build a Flask + HTML dashboard for experiment results via @viz-frontend.
Instructions
When this skill is invoked with optional <argument>:
Step 1: Determine target
- If argument is
overview → project overview mode (delegates to @slides-maker Mode 3)
- If argument is
compare → comparison mode (cross-experiment)
- If argument is an exp ID (e.g.,
exp01a) → that experiment
- If no argument → read
.pipeline-state.json for current_exp
- If
current_exp is null and no argument → ask user which experiment or overview
Step 2: Gather data
For single experiment:
- Read
exp/{exp_id}/results/summary.md for quantitative results
- Read
exp/{exp_id}/README.md for experiment context
- Check if
viewer/app.py already exists (update vs create)
For compare mode:
- Read
exp/summary.md for cross-experiment overview
Step 3: Delegate to @viz-frontend
For single experiment:
Build a results dashboard for experiment {exp_id}.
Results data: {summary.md content}
Experiment context: {README.md content}
Viewer directory: viewer/
If viewer/app.py exists, add/update the view for this experiment. Otherwise create from scratch.
For compare mode:
Build a cross-experiment comparison dashboard.
Summary data: {exp/summary.md content}
Viewer directory: viewer/
For overview mode:
- Ask user what type: "slides" (default), "onboarding", or "demo-script"
- Delegate to @slides-maker:
mode: overview
content_type: {user's choice}
Step 4: Report
Tell user: "Dashboard ready. Run python viewer/app.py and open http://localhost:5001"