Concept: Onboarding — proactively offer to import existing skills (e.g. from Claude) for a fast first success

[iret77]

Part of epic #391 (skill import from other AI systems). This is Part 3 of that
concept, split out for separate tracking. It builds on the import mechanism (Part 1) and
the skill-aware Builder (Part 2) described in #391.

Summary

Once a user can import skills (epic #391), the remaining onboarding piece is to make
the offer proactively: during new-user setup, invite the user to bring a few skills they
already trust — first from Claude (Agent Skills), later from OpenAI (Codex / ChatGPT) —
so they reach an early "it works, with my stuff" moment instead of staring at an empty
workspace.

This is a concept, not an implementation plan.

Why

First impressions decide adoption. A new operator who lands in an empty workspace has to
imagine what omadia can do; a new operator who imports a skill they wrote last week and
watches an agent use it has already succeeded. Reusing assets the user already trusts is
the shortest path to that first success — and it directly reduces the cold-start tax called
out in the broader onboarding work (#337, #338).

What the user sees (UX sketch)

1. In the first-run / setup flow, an optional, skippable step: "Bring your skills."
2. The user drops one or more SKILL.md files (or pastes). omadia previews each
   (name / description from frontmatter) — no size or "slot" talk, ever.
3. On confirm, the skills are created in the registry via the same import path as Concept: Skill import from other AI systems (Claude, Codex/ChatGPT) + effortless skill UX in the Agent Builder #391
   (no parallel mechanism).
4. omadia offers to spin up a starter agent that already uses one of them, so the user
   has a working, personal example in the first minutes — not a blank canvas.
5. The step is fully skippable for users who want to start clean.

Principles

• Optional and skippable — never a gate in front of the product.
• Fast path to a working agent — the payoff is a running agent using the user's own
  skill, not just a registry full of imported entries.
• One import path — reuse Concept: Skill import from other AI systems (Claude, Codex/ChatGPT) + effortless skill UX in the Agent Builder #391's mechanism and the skill-aware Builder; do not build a
  separate onboarding-only importer.
• Untrusted by default — same preview-before-activate and guard posture as any imported
  content.

Feasibility anchors in omadia

• The setup wizard / onboarding flow already exists as the place to add an optional
  step.
• The import mechanism and skills registry come from epic Concept: Skill import from other AI systems (Claude, Codex/ChatGPT) + effortless skill UX in the Agent Builder #391 — this issue is the
  onboarding surface on top of them, not new core machinery.
• Starter / template galleries already exist in the builder surface and can seed the
  "working example" agent.

Non-goals

• The import mechanism itself (Concept: Skill import from other AI systems (Claude, Codex/ChatGPT) + effortless skill UX in the Agent Builder #391, Part 1) and the skill-aware Builder (Concept: Skill import from other AI systems (Claude, Codex/ChatGPT) + effortless skill UX in the Agent Builder #391, Part 2).
• Any automated connector that pulls skills out of Claude/OpenAI accounts — manual upload /
  paste is the starting point; a connector is a separate, later question.
• Bulk migration tooling.

Open questions

• Where exactly in the first-run flow does the offer sit, and how is it phrased so it reads
  as a quick win rather than a chore?
• How many skills do we suggest importing for a good first run (one, a few)?
• Do we auto-create a demo agent that uses an imported skill, or only offer it?
• Sourcing: manual upload first — is there appetite for a future Claude/OpenAI connector,
  or is upload/paste sufficient?

Related

• Concept: Skill import from other AI systems (Claude, Codex/ChatGPT) + effortless skill UX in the Agent Builder #391 — epic: Skill import + skill-aware Builder (Parts 1–2)
• Concept: setup-prompt onboarding — agent-driven local install via a public skill file #338 — Concept: setup-prompt onboarding (agent-driven local install via a public skill file)
• Lower the barrier to try omadia (no-Docker eval path) #337 — Lower the barrier to try omadia