The Adversarial Checklist

Before trusting any solution, attack it with these five shapes:

1. All ties. Every element equal; every weight identical; every deadline the same. Tie-break branches that never ran on random data run now.
2. Asymmetry. A path graph; one huge element among tiny ones; an interval swallowing all others. Symmetric data flatters symmetric bugs.
3. Degeneracy. n = 1; n = 0 if legal; empty ranges; single-point intervals; self-loops; duplicates everywhere.
4. Extremes. Maximum values (overflow), minimum values (underflow of assumptions like "positive"), alternating max/min patterns that maximize comparison counts.
5. Adversarial order. Sorted-descending input to a "sort then process" claim; reversed trees; queries ordered worst-case-first.

The mindset

For each shape, the question is not "will my code break?" but "what would I input if I were trying to break it?" That inversion finds bugs that rereading never will, because you wrote the code and your brain auto-corrects while reading.

Cost-benefit

The checklist costs minutes. A wrong submission costs the problem, the confidence, and often the contest. There is no scenario where skipping the checklist is correct arithmetic.