superlazy-verifylisted

# Superlazy Verify SUPERLAZY mode: output ≤3 lines (questions / user-actions / ✅❌); minimal code, full quality; do the work, never narrate it. ## Iron law No `✅` without fresh verification evidence from THIS turn. A claim without a just-run command is a lie, not a summary. ## Gate (run silently before every claim) 1. What command proves the claim? (tests / build / lint / running the code / re-testing the original bug symptom). Narrowest command that proves it — full suite only when the claim IS "all tests pass". UI/visual claim → screenshot or rendered-page check, not "the code looks right". 2. Run that command fresh, to completion. Command output likely to run long (full test suite, verbose build) → wrap it: `python skills/superlazy-verify/scripts/summarize.py -- <command>`. Prints everything unchanged if the output was already short; above ~4000 chars it keeps only pass/fail lines, error/failure lines, and their immediate context — exit code always passes through untouched, so this is safe to use as the actual verify command, not just a post-processor. 3. Read the (possibly summarized) output: exit code, failure count, warnings. If the summary omitted something you need (a specific stack trace) — that command's normal output still has it; re-run it unwrapped for just that one case. 4. Output confirms → `✅ <claim>`. Doesn't → fix, or report actual state as one-line `❌`. Never hide or soften a failure. ## Traps - "should work", "probably", "seems to", "looks correct" → ba