Did the work actually land? The honest track record.
Every AI tool stops at the draft โ nobody ever learns whether the PRD shipped, the memo got approved, or the cold email got a reply. The Ledger closes that loop: log what you made, then come back and mark what actually happened. Over time it becomes the one thing no skill library has โ a record of which skills' outputs land in the real world.
Landed-rate = shipped/approved/got-a-reply รท resolved outcomes. Needs a few marked entries to mean anything.
Contributing sends only skill โ counts (no titles, no content) so the community can build a public "what lands" leaderboard โ the honest signal that feeds improve-skill.