naturalizelisted

# Naturalize Rewrite text so it reads like the user actually wrote it, not like an LLM drafted it — grounded in a profile calibrated once from the user's own writing (not guessed from a generic "sound more human" instruction), then reused silently on every later request until the user asks to change it. `<SKILL_DIR>` below is this skill's own announced base directory (the path shown when this skill loads, ending in `.../skills/naturalize`) — references live at `<SKILL_DIR>/references/`. ## Step 1 — Determine request mode Default behavior is `profile` mode: the saved profile (if any) governs everything. The user can override this for a single request by explicitly asking to skip it — "naturalize this generically," "without using my profile," "ignore my profile for this one," "just de-AI-ify this, don't personalize it." That sets `explicit_generic` mode for this request only. - **`profile` mode** (default) → go to Step 2. - **`explicit_generic` mode** → **do not read or write `~/.naturalize/profile.md` at all** for this request — not even to check whether it exists. Skip straight to Step 5, using the checklist's stated defaults (`references/checklist.md`): everything on except typos, no custom instructions, no voice/sample matching. This is a one-off override — it never triggers calibration and never modifies a saved profile. If it's ambiguous whether the user means this request only or wants to change their standing default, ask; don't silently reinterpret a