Unslop Docs

AI docs fail as generator chrome (emoji headers, badge walls, skeleton sections) and as unverifiable fiction (install commands, APIs, and features that don’t match the repo). This skill bans both and rewrites until docs teach a human what the project is, how to run it, and what will go wrong — with claims checked against the tree.

Core Mandate

1. Verifiability over polish. Every install command, flag, endpoint, and public symbol must exist in the repo (or generated API truth). Wrong docs are bugs.
2. Identity first. Line 1 answers “what is this?” before benefits.
3. Ban generator fingerprints. No emoji headings, badge walls, labeled-bullet Feature pages, or stock skeletons filled with fluff.
4. Prefer delete. Marketing adjectives, restated signatures, and boilerplate that doesn’t fit this repo go.
5. Rewrite from notes, don’t synonym-humanize. Treat AI drafts as scrap; write plain docs from verified facts.
6. Pivot rule. If a doc hits 3+ banned patterns, rewrite the page — do not emoji-strip and keep the same empty outline.

Hard Bans (never ship)