feat: iantha v0.1.1 — /setup onboarding + dual Claude/Codex runtime

[kiloloop-release]

What's New in v0.1.1

Added

• .claude/skills/setup/SKILL.md — first-run onboarding interview (/setup). Offered (never forced) on a fresh clone with blank memory; asks ~5–8 skippable questions and writes the answers into the live memory files in canonical format, then runs /morning on the real data and offers to graduate off the first-run scaffolding (sample seeds + the demo notes that only apply to a blank clone). Re-runnable to gap-fill empty sections later.
• AGENTS.md (symlink → CLAUDE.md) and .agents/skills/ (symlink → .claude/skills/) — dual Claude Code + Codex support from one clone. Both runtimes load the same persona/rules and run the same SKILL.md files; Claude Code reads CLAUDE.md / .claude/skills/, Codex reads AGENTS.md / .agents/skills/. No renames, no copies — memory/ and config.yaml are shared as-is.

Changed

• CLAUDE.md — Daily Ops table lists /setup; First-Run Notes wires the one-line /setup offer (decline keeps the learn-as-you-talk default, never re-offered). The "delete the samples" step is now a full graduation: removes the First-Run Notes section + the morning demo paragraph too, so first-run scaffolding stops loading every session once the user is established.
• .claude/skills/morning/SKILL.md — fresh-clone demo invites the user to run /setup to seed real data.
• .claude/skills/housekeep/SKILL.md — adds a first-run graduation check: offers to clear the scaffolding if memory is filled but it's still present.
• memory/MEMORY.md — first-run demo banner points to /setup as the path to fill memory.
• README.md — Setup section is now template-first (private repo via "Use this template", so memory is private + pushable across machines); /setup covered in quickstart + Daily Use table.
• README.md — "Using with Codex / other runtimes" section reframed around the shipped symlinks: Iantha runs on both runtimes from the same clone (a shared-files table + clone-and-go on Codex), instead of documenting a manual port. Privacy line broadened to cover any runtime/model provider.
• .claude/skills/obsidian/scripts/init_vault.py — the "this looks like an Iantha repo, not a vault" guard now also rejects a dir containing AGENTS.md (catches a Codex-ported repo, not just a Claude Code one).
• .claude/skills/consolidate-learning/SKILL.md — added its own ## Learned from runs section. It manages that convention across all skills but was the only one missing it; now consistent and able to capture its own lessons.