Podium

A Claude Code / Claude agent skill that scores an idea the way a competition judge actually scores it.

Most competition advice tells you how to build a better product. Judges don't score products — they score four rubric columns, in a room, under time pressure, against two dozen other teams, and then they hand out prizes that reward different things from those columns. Podium scores the way the room actually scores.

Use it for hackathons, science fairs, pitch competitions, innovation awards, startup contests, demo days, grant rounds, and school or university competitions — before you submit.

What it does

• Incumbent check first. Runs a real search for who already solves the problem, because that's the most common way a good idea dies, and it dies late — in Q&A.
• Scores the four columns (creativity, practicality, impact, presentation) in honest bands, with reasons.
• Rehearses the five questions a team has to survive, and names which column each unanswered one drains.
• Flags the four quiet killers — permission on the critical path, exposure numbers posing as outcome numbers, a demo that must be simulated, a heavy dependency carried into an adaptation round.
• Names the prize you're actually playing for. Judge-scored places, impact awards and crowd votes reward different things, and teams routinely optimise for the wrong one.
• Pattern-matches against projects that really won — see references/winners.md, a set of verified entries from real competitions with the shape each one demonstrates.

It is calibrated to say "mid" when an entry is mid. A skill that predicts a win every time is a horoscope.

Install

git clone https://github.com/Creeperprol/podium-skill.git ~/.claude/skills/podium

Then just ask, in any competition context: "will this win?", "how would judges score this?", "what should I fix before I submit?"

Layout

SKILL.md              the scoring method
references/winners.md verified winners and the shapes that win and lose

License

MIT