Hackathon Judge — four-pass project scoring

Score a project submission with the methodology from Don't Let the LLM Pick a Number: independent passes collect bounded signed evidence; math computes the scores; a final pass turns the evidence into mentoring feedback for the team.

This skill is generic by design — there is nothing hackathon-specific in the math. "Hackathon" is the natural search term, but the same pipeline works for take-home interviews, demo-day judging, internal review boards, vendor pilots, and any project where someone built a thing and someone else needs to score it.

When to use this

• A submission has both a codebase and an optional demo video (URL or transcript) — the four-pass shape gets the most value when both exist, but the skill degrades gracefully if only code is available.
• The user wants a defensible numeric score per criterion, not a vibe verdict.
• Multiple submissions need to be scored consistently (the formula is reproducible across runs and reviewers).
• The user wants the team to receive grounded mentoring feedback, not a leaderboard rank.

If the user just wants quick feedback on a draft (no codebase), use what-works-feedback-judge. If they want to design their own scoring rubric for a non-project domain, use evidence-scoring.

The four passes

Each pass is a separate stage. They can be run together (full pipeline) or in isolation when iterating.