auto-improve round 3: ruff S603/S607 are intentional (subprocess is our core function). Added pipeline JSON files.

auto-improve round 2: apply 2 debate fixes (debate.py)

auto-improve round 1: fix 5 ruff S110 findings (silent except:pass)

feat: add focus instruction + round count to /multi-auto

auto-improve round 1: apply 4 suggestions from self-review

feat: add /multi-auto — self-improving loop with engine scout + Claude surgeon

fix: stderr for progress output in debate.py and pipeline.py (keeps JSON clean)

feat: add /multi-debate, /multi-pipeline, /multi-ui — conversation, custom chains, UI analysis

feat: add Tier 1 commands — /multi-security, /multi-estimate, /multi-compare, /multi-tradeoff

docs: evaluate 100 potential commands — 25 worth it, 40 one-model-fine, 35 not-worth-it

feat: add /multi-doc and /multi-plan commands

feat: add /multi-debug command — multi-model debugging with hypothesis consensus

feat: add /multi-test command — multi-model test generation

feat: round 3 — fix self-review findings

feat: round 2 — fix review finding + apply improve suggestions