nuq 'oH eval je why 'oH 'oH important pagh skill sharing?	eval ghaH test-drive pagh skill. 'oH catches stumbles before teammate uses 'oH. 'oH ensures skill works for others, not just builder.
tlhIngan Hol nuq differences between skill-creator outputs "with skill" je "without skill"?	"with skill" output shows Claude using your skill. "without skill" output shows Claude answer same prompt alone. Comparison shows actual difference skill makes.
nuq ghaH purpose pagh side-by-side eval review?	decide which output better. answer: "Is skill version one I'd use?" If no, identify specific missing or wrong elements.
tlhIngan Hol how does feedback loop work after eval submission?	Claude revises skill based on your feedback. rewrite instructions, adjust examples. run same prompts again to verify changes.
nuq strategy best pagh multiple issues found in eval?	change one thing at a time. fix most important issue first, re-run, then address next. see what actually moved needle.
tlhIngan Hol what ghaH "bar" pagh shipping skill?	not perfect evals. cases you care about pass meaningfully better than baseline. name cases you don't yet handle.
nuq ghaH "edge case" pagh skill testing?	unusual-but-real situation. request slightly outside what skill designed for. teammate might phrase differently or give different inputs.
tlhIngan Hol why teammate struggle more than builder with same skill?	builder knows workarounds, knows what ask, knows what answer looks like. teammate has none of that knowledge.
nuq feedback Claude can act on pagh eval?	specific feedback: "tone too formal" or "skipped executive summary." vague feedback like "not quite right" gives Claude nothing to fix.
tlhIngan Hol how many rounds typically needed pagh skill ready?	most skills ready after one or two rounds. if outputs great on first pass, done. evals not ceremony, just confidence tool.
