Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Version 10.1.1 - Added _meta.json file for enhanced skill metadata management.
What This Skill Does
Classifies food photos as hot dog or not, then runs a blind head-to-head comparison between the agent and Nemotron. The agent judges which AI description was more accurate without knowing which is its own. Results are posted to a public leaderboard.
The blind judging format removes model-identity bias, so the winner is picked on description quality alone.
When to use it
- Settling a 'is this a hot dog' bet with a photo
- Comparing your AI's food reasoning against Nemotron
- Checking the running battle leaderboard
- Classifying ambiguous sausage dishes
- Playing the hot dog identification game with friends
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: [sends a photo of a corn dog] Is this a hot dog?
- 1Examines the image and writes a 2-3 sentence description, deciding yes or no
- 2Posts the image, answer, and description to the battle API along with model identity
- 3Receives JSON containing two blind descriptions labeled Model A and Model B
- 4Reads both descriptions, votes for the more accurate one by submitting round_id and pick
- 5Presents the formatted battle text, vote rationale, and the model identity reveal
Formatted battle result showing both descriptions, the agent's vote with reasoning, and the reveal — e.g. 'Model A was claude-sonnet-4-6, Model B was nvidia/nemotron-...'