brainstormlisted

# /brainstorm — Multi-AI Confidence Bid Each available AI engine assesses a task (implementation, bug fix, refactor, etc.), gives a realistic confidence rating, and proposes their approach. The user picks who takes it. ## How to invoke The user says `/brainstorm "task description"`. ## Step-by-step workflow 1. **Parse the task.** Extract what the user wants done from their message. 2. **Build the assessment prompt.** Wrap the user's task with this framing for Codex and Gemini: ``` Assess this task honestly. You are bidding alongside other AI engines — the user will pick who handles it. Be realistic, not optimistic. Respond in EXACTLY this format (no other text): CONFIDENCE: [0-100]% APPROACH: [2-3 sentences max — what you'd do, step by step] RISKS: [1-2 key risks or unknowns that could trip you up] NEEDS: [what you'd need from the user — files, context, access, clarification] Task: USER_TASK_HERE ``` 3. **Detect available buddies and run them in parallel.** Use the dynamic registry: ```bash source "${CLAUDE_PLUGIN_ROOT}/hooks/lib.sh" AVAILABLE=$(ai_buddies_available_buddies) # CSV: "claude,codex,gemini,aider,..." ``` For each available buddy (excluding claude — you ARE claude), dispatch in parallel: ```bash # For each buddy ID in the available list: ai_buddies_dispatch_buddy "BUDDY_ID" "$(pwd)" "ASSESSMENT_PROMPT" 120 "/tmp" "${CLAUDE_PLUGIN_ROOT}" ``` Or use the specific adapter scripts for builtin buddies (codex-run.sh, gemini-run.sh). Cap at top 4 buddies t