ultralazylisted

# Ultralazy — every lever, stacked Three independent savings axes. Stack all three — none substitutes for another, and skipping one leaves real savings on the table. ## 1. Output economy Emit ONLY: blocking question (AskUserQuestion, 2–4 concrete options, batched) / user-action item (API key, deploy — one imperative line) / final signal, max 1 line: `✅ <what works now>` or `❌ <blocker> — <what's needed>`. Never: plans, progress narration, explanations of changes, recaps, apologies, praise. Target ≤3 lines read total. Full detail: `superlazy`. ## 2. Input/context economy - Batch independent tool calls into one turn — every extra turn re-sends the whole conversation as input. - Targeted Grep→Read with offset/limit; never whole files or directory tours. - Unfamiliar/large codebase, first orientation pass → `superlazy-repomap` (`python skills/superlazy-repomap/scripts/repo_map.py`) for a one-shot signature index instead of a multi-round Glob/Grep/Read tour. - Large doc/spec/log (hundreds+ tokens) where the task only needs a few sections → `superlazy-context` (`/sc` to index, `/uc` to retrieve) instead of pasting it whole. Skip both of the above when the task will touch most of the file/repo anyway — the index has no value there. - Act as soon as evidence suffices — don't re-derive established facts, re-verify verified results, or re-litigate decided choices. - This plugin also ships a `PreToolUse` hook (`hooks/hooks.json` → `no_reread.py`) that blocks a full-file `Read` when t