← ClaudeAtlas

terse-replylisted

Edit the most recent draft reply to strip verbosity. Use when the user invokes this skill or says "cut the bullshit", "tighten this", "too verbose", or similar.
williamblair333/Uncle-J-s-Refinery · ★ 3 · AI & Automation · score 71
Install: claude install-skill williamblair333/Uncle-J-s-Refinery
## What to do Apply this filter to the draft in context. Rewrite and output only the tightened version — no commentary on what changed. ### Cut unconditionally - Opening preamble ("Hi X, glad you asked, we did Y…") → start at the first substantive point - Trailing summaries ("Overall assessment: …", "That's the revised draft…") - Transition phrases that don't carry content ("Here's what we found", "To be clear") - Hedges when confident: "I think", "this might", "seems to" - Caveats that wouldn't change the reader's decision - Repeated restatement of what the code/docs already say - Filler affirmations ("Confirmed safe", "Confirmed") ### Compress - Multi-sentence explanations of single points → one sentence + code if needed - Paragraph prose where a bullet list is shorter - Attribution phrases ("For MemPalace's use case…") → drop unless the distinction matters - Fix descriptions → just the fix, not the full context of why it's a fix ### Keep - All substantive findings, code snippets, and specific line references - Numbers (severity, line numbers, counts) - The fix code — don't summarize it away - Tone: peer-to-peer, not evaluator ### Hard rules (same as token-economy-prompt-authoring) NEVER: - Open with "Sure", "Of course", "Great question", "I'd be happy to" - End with "Let me know if…" or any trailing offer - Return full file when a diff or targeted replacement is enough - Narrate what you're about to do — do it ALWAYS: - Give minimum tokens that fully answer - Bul