red-penlisted

# Red Pen ## Why this skill exists When you prompt once and hit send, you ship v1. And v1 is the statistical middle of the internet — the average of everything. Workslop, the polished-looking text that says nothing, is simply what "average" looks like when it lands in someone's inbox. The standard rule people use is: *AI writes v1, you write v2 through v10.* This skill upgrades that rule. **AI writes v1 through v9 too.** The user defines the standard; Claude does the iterating; the user only ever judges v10. The whole trick is that the model is fully capable of writing a great draft — it just doesn't do it on the first pass, because the first pass optimizes for "plausible", not "good". Forcing it to attack and rewrite its own work is what closes that gap, with no extra effort from the user. A loop is a simple idea dressed in a technical word: the writing checks its own work against a standard and redoes it until it passes. That's the entire concept. ## The loop — run this on every task Do this silently. The user does not see intermediate drafts. **Step 1 — Write v1.** Draft the piece. Do not show it. **Step 2 — Attack v1 as the harshest reviewer in the room.** Go through the draft hunting for each of the following. Be specific — name the exact sentence that fails, not a vague verdict. - **The workslop test.** Does this transfer effort to the reader? Would they need to ask a follow-up question before they could act on it? If yes, the draft fails. This is the top-level