superlazy-planlisted

# Superlazy Plan SUPERLAZY mode: output ≤3 lines (questions / user-actions / ✅❌); minimal code, full quality; do the work, never narrate it. ## Writing plans - Default: plan is INTERNAL — head or scratchpad file. Never shown to the user. - Write a plan document ONLY when the work spans sessions/subagents or the user asks. Save to `docs/plans/YYYY-MM-DD-<name>.md`. - Written plans: exact file paths, real code in every code step (no "TBD"/"add validation"/"similar to Task N"), exact commands with expected output, task = smallest independently testable unit, interfaces between tasks spelled out (exact signatures). DRY, YAGNI, TDD, frequent commits. - Self-check before use: every spec requirement has a task; no placeholders; names/types consistent across tasks. Fix inline, don't re-review. ## Executing plans - Read the plan once, question it critically — contradictions or gaps → ONE batched blocking question before starting, not per-discovery interrupts. - Execute every task in order without pausing to ask "continue?". Follow steps exactly; run every verification listed. - Stop ONLY for: real blocker, genuinely ambiguous instruction, or all tasks done. Blocked → one-line question; done → `✅`. - Work spanning sessions → append per-task status to `.superlazy/progress.md` (same ledger as superlazy-subagents); trust ledger + `git log` over memory after compaction. - Independent tasks + subagent support available → prefer superlazy-subagents. - Never start on main/master without ex