v4.1.0 — any-agent brain

Minor. The environment generalizes: any AI CLI (Claude Code, Codex, Gemini, Cursor) can be the brain. kj solomon — a neutral third AI arbitrates brain↔reviewer conflicts (security findings are never arbitrable). kj brief <role> — outcome-first role briefs the brain absorbs. kj agent run — the inter-agent bus. kj report-issue — sanitized, deduped, human-gated upstream reporting: the ecosystem repairs itself. Playbook targets gemini/all, kj init --json, and new five-page v4 docs (EN+ES) with v3 grouped as legacy. Full notes in the CHANGELOG.