[Feature] Onboarding completion criteria — at least one skill or exhausted attempts

[graycyrus]

Summary

Define and enforce clear completion criteria for the onboarding conversation: agent marks done only after user connects at least one skill OR agent has genuinely tried and user declined.

Problem

Currently the welcome agent can call complete_onboarding({"action": "complete"}) at any point with no enforced minimum engagement. This means onboarding can be "completed" after a single exchange with no skills connected.

Solution

Combine prompt-level and code-level guards:

Prompt-level (in prompt.md):

• Agent instructed to only call complete after: (a) at least one skill connected, OR (b) offered connections 2-3 times and user explicitly declined
• AND at least 3 back-and-forth exchanges have occurred

Code-level (in complete_onboarding.rs):

• complete action validates minimum criteria before flipping the flag
• If criteria not met, return error with reason: "User hasn't connected any skills and minimum exchanges not reached"
• The check_status action returns ready_to_complete: bool and reason: string

Graceful fallback:

• If user says "skip", "later", "not now" — agent acknowledges, briefly explains how to connect later (settings), then completes
• If user is clearly disengaged after 3+ exchanges — agent wraps up gracefully

Acceptance criteria

• Minimum exchanges — complete action rejects if fewer than 3 back-and-forth exchanges
• Skill attempt — Agent tries to get at least one skill connected before completing
• Graceful decline — If user declines all skills, agent completes after explaining how to connect later
• Skip support — "Skip" / "later" / "not now" recognized as explicit decline, handled gracefully
• Server-side enforcement — complete_onboarding tool enforces criteria, not just prompt-level
• Status reporting — check_status returns ready_to_complete with reason

Related

• Files: src/openhuman/tools/impl/agent/complete_onboarding.rs, src/openhuman/agent/agents/welcome/prompt.md
• Depends on: [Bug] isOnboarded (chat_onboarding_completed) flag set too early #591, [Feature] Rewrite welcome agent system prompt for conversational onboarding #595
• Part of: Conversational onboarding agent epic

---

Parent: #599 — Conversational onboarding agent