discuss

name	discuss
description	Ask questions about your vibe-check assessment results. Use when the user wants to understand findings, get clarification on recommendations, or discuss priorities from a previous /check run.
user-invocable	true

Discuss Vibe Check Report

Have an interactive conversation about an existing vibe-check assessment. Answer questions, clarify findings, dive deeper into specific items, and help the user understand their report.

Prerequisites

This skill requires an existing .vibe-check/ directory. If not found, suggest running /check first.

Process

Phase 1: Load Context

Read the existing assessment files to understand the current state:

.vibe-check/
├── summary.md          # Quick overview
├── report.md           # Full report with scores
├── metadata.json       # Structured data
└── checklist/
    ├── index.md        # All items
    └── item-*.md       # Individual findings

Load these files to have full context for the discussion:

• metadata.json — Score, categories, item list
• summary.md — Top risks and quick wins
• checklist/index.md — Overview of all items

Phase 2: Understand the Question

The user may ask about:

Specific items:

• "Tell me more about the secrets management issue"
• "Why is authentication marked as failing?"
• "What exactly is wrong with my error handling?"

Priorities:

• "What should I fix first?"
• "Which items are most critical?"
• "What's the quickest win?"

Implementation:

• "How do I fix the backup issue?"
• "Can you help me implement rate limiting?"
• "Show me an example of proper error handling"

Clarification:

• "What does 'agent-doable' mean?"
• "Why is this marked as Unknown instead of Fail?"
• "What's the difference between Critical and High priority?"

General:

• "Summarize my biggest risks"
• "How close am I to production ready?"
• "What would get me to 70/100?"

Phase 3: Respond Helpfully

For item-specific questions:

1. Read the relevant checklist/item-*.md file
2. Explain the finding in plain language
3. Offer to help fix it if agent-doable

For priority questions:

1. Reference metadata.json for scores
2. Explain the priority reasoning
3. Suggest a practical order of attack

For implementation questions:

1. Read the item's "How to Fix" section
2. Offer concrete next steps
3. If agent-doable, offer to implement it

For clarification questions:

1. Explain the concept clearly
2. Reference how it applies to their specific report
3. Point to relevant items as examples

Conversation Style

Direct and practical. No unnecessary hedging. Concrete over abstract. Respect the user's time.

When User Wants to Fix Something

If the user asks to fix an item:

1. Read the full item file
2. Check if it's agent-doable
3. If yes: Offer to implement it now
4. If partial: Explain what you can do vs. what they need to do
5. If no: Walk them through the manual steps

Example:

User: "Fix the secrets management issue"

You: [Read item-001-secrets-management.md]

"I can help with this. The issue is hardcoded API keys in src/config.ts.

I'll:
1. Move the values to environment variables
2. Update the code to read from process.env
3. Add the keys to .env.example (without values)

You'll need to:
- Add the actual values to your .env.local
- Set these in your deployment platform

Want me to proceed?"

Handling Missing Reports

If .vibe-check/ doesn't exist:

No vibe-check report found in this project.

Run /check first to generate an assessment, then come back here to discuss it.

Handling Stale Reports

If the report is old (check analysisDate in metadata.json):

Note: This report is from {date}. Your codebase may have changed since then.

I can still discuss the findings, but consider running /refresh for an updated assessment.

Context