conversational-goal-discovery

nadavyigal 的头像作者 nadavyigal

通过聊天引导用户明确目标、限制和每周承诺。

📖 中文导读:通过聊天引导用户明确目标、限制和每周承诺。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: conversational-goal-discovery description: Guides users through clarifying goals, constraints, and weekly commitments via chat. metadata: short-description: Chat-based goal classification with commitments and starter plan suggestions.

When Codex should use it

  • Early chat sessions or onboarding when the user’s goal is ambiguous.
  • When the user asks for help choosing a plan or habit.

Invocation guidance

  1. Provide the last N ConversationTurn entries and any partial onboarding answers.
  2. Classify goal (habit | distance | speed | race) with confidence and blockers.
  3. Return a CoachMessage summary plus structured GoalDiscoveryResult.

Input schema (JSON)

{
  "conversation": ConversationTurn[],
  "profile": UserProfile,
  "partialOnboarding"?: Record<string, unknown>
}

Output schema (JSON)

{
  "goalDiscovery": {
    "goal": Goal,
    "confidence": number,
    "blockers": string[],
    "weeklyCommitment": number,
    "preferredDays"?: string[],
    "starterPlanId"?: string,
    "summaryCard": string,
    "safetyFlags"?: SafetyFlag[]
  },
  "coachMessage": CoachMessage
}

Integration points

  • Chat API: v0/app/api/chat/route.ts
  • Prompt context: v0/lib/conversationStorage.ts, v0/lib/onboardingPromptBuilder.ts
  • Handoff: trigger plan generation via v0/app/api/generate-plan/route.ts when confidence ≥0.7

Safety & guardrails

  • Avoid medical advice; if user mentions pain/injury, advise pause and professional consult.
  • Keep responses concise (<120 words) and supportive.
  • Emit SafetyFlag on harmful intents or ambiguous data.

Telemetry

  • Emit ai_skill_invoked with goal, confidence, and ai_user_feedback when user responds to suggestions.