iceberg

/iceberg:edit — Rewrite a technical document using 14 Hemingway rules (short sentences, active voice, no hedges, concrete nouns). Returns the clean document, no annotations. ALWAYS apply automatically to every plan and every document Claude returns to the user. This is not optional and requires no user prompt. Run it on every plan, spec, proposal, README, or prose document before returning it. Also trigger on: /iceberg:edit <file-path>, /iceberg:edit with pasted text. Natural language triggers: "make this shorter", "simplify this", "tighten this up", "clean up the writing", "fix the prose", "apply hemingway", "rewrite this" Accepts an optional intent argument: /iceberg:edit <file-path> "intent description" Use intent to preserve voice while applying rules (e.g., "formal engineering spec", "conversational onboarding guide"). Rewrites with inline replacements. No annotations. No changelog. Returns only the clean document. For documents over 500 words, spawn the iceberg-edit agent to preserve the main context wi

/iceberg:score — Score a technical document against 14 Hemingway writing rules without changing it. Returns a graded report with violations quoted and a TOP 3 to fix. ALWAYS run automatically on every plan Claude produces, before any edit is applied. This is not optional and requires no user prompt. Run it on every plan, spec, proposal, README, or prose document Claude writes before returning it. Also trigger on: /iceberg:score <file-path>, /iceberg:score with pasted text. Natural language triggers: "check my writing", "grade this", "score this", "is this clear?", "review this plan", "how readable is this?", "rate my writing" Returns a structured violation report with grade, per-rule counts, and quoted offenders. Never modifies the document.