Quick Question Command

Answer questions using the most appropriate MetaSaver agent without building anything.

Use when: You have a question that needs agent expertise but doesn't require code changes.

---

Entry Handling

When /qq is invoked, ALWAYS proceed to Phase 1 regardless of prompt content. Questions are analyzed first, then routed to the appropriate agent for answering.

---

Phase 1: Analysis (PARALLEL)

Follow: /skill scope-check, /skill agent-check

Spawn 2 agents in parallel to execute analysis skills. Collect: scope (with targets and references), recommended_agent

---

Phase 2: Agent Selection

Follow: /skill agent-selection

Validate recommended agent exists in MetaSaver agent pool. Fallback to code-explorer if agent not found.

---

Phase 3: Question Execution

Follow: Selected agent from Phase 1 or Phase 2

Spawn MetaSaver agent in "answer mode":

• Agent receives original question + scope context
• Agent researches using tools (Serena, Grep, Read)
• Agent analyzes codebase if needed
• Agent formulates comprehensive answer
• Provide answers inline only (read-only mode)

---

Phase 4: Return Answer

Agent returns answer to user with:

• Direct answer to the question
• Supporting evidence (file paths, code snippets)
• References for further reading if applicable

---

Examples

/qq "How does authentication work in this app?"
→ Analysis → code-explorer → Answer with auth flow details

/qq "What validation patterns do we use for forms?"
→ Analysis → coder → Answer with form validation patterns + examples

/qq "Why is ESLint configured this way?"
→ Analysis → eslint-agent → Answer with ESLint config explanation

/qq "How does rugby-crm handle user roles compared to here?"
→ Analysis → code-explorer → Answer comparing role handling across repos

/qq "What's the database schema for users?"
→ Analysis → prisma-database-agent → Answer with schema details

/qq "What's our testing strategy for API endpoints?"
→ Analysis → tester → Answer with testing patterns + coverage

---

Enforcement

1. Use AskUserQuestion tool for every question to the user. Present structured options with clear descriptions.
2. ALWAYS run Analysis phase first (scope + agent check in PARALLEL)
3. Run analysis skills in PARALLEL (single message, 2 skill calls)
4. ALWAYS validate agent exists before spawning
5. Use code-explorer as fallback if agent not found
6. Agent executes in "answer mode" - read and analyze only
7. Answer questions directly - skip PRD creation
8. Provide immediate responses - skip approval gates
9. Focus on answering - skip execution planning
10. Include file references in answers when relevant
11. Keep answers focused and concise
12. Respect scope boundaries from scope-check