research-councillisted

# `/research-council` — Expert Research Council Launch a panel of specialized AI experts to research a topic in depth. Each expert searches the web, reads your code, and produces a structured report. A merger agent consolidates everything into actionable recommendations. **Why a panel instead of one long prompt:** a single agent asked to "consider all angles" produces one voice wearing several hats, and it converges on its own first idea. Separate agents with separate contexts genuinely disagree — and the disagreements are the valuable part. Keep them in the output; do not average them away. ## Modes ### `new <topic>` — full research from scratch 1. Select the experts relevant to the topic (not all of them — see below). 2. Each expert: web research + code analysis + structured report. 3. Merge all reports into a consensus view. 4. Output: consolidated report + implementation plan. ### `analyze <path>` — analyze existing research documents 1. Read every file in the given path. 2. Score each document (1-100). 3. Extract the key ideas, discard the filler. 4. Merge into consensus + plan. ### `merge <report1> <report2> …` — merge specific reports ### `plan <consolidated_report>` — generate an implementation plan from a report ## Expert panel Pick the experts that fit the question. A panel of four sharp, relevant experts beats ten where six are padding — every irrelevant expert adds confident noise the merger has to filter. | Expert | Persona | Focus | Tools | |---|---|---