- name
SKILL.md
El archivo de la skill permanece en el repositorio del publicador.
Pega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add K-Dense-AI/scientific-agent-skills --skill consciousness-council -a cursor -yPega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add K-Dense-AI/scientific-agent-skills --skill consciousness-council -a claude-code -yPega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add K-Dense-AI/scientific-agent-skills --skill consciousness-council -a codex -yPega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add K-Dense-AI/scientific-agent-skills --skill consciousness-council -a grok -yEnvía el mensaje copiado en un chat de Grok Bot. El Bot obtiene este SKILL.md de GitHub y lo añade a las skills que comparten todos tus Bots.
Install this skill: https://github.com/K-Dense-AI/scientific-agent-skills/tree/HEAD/skills/consciousness-councilDel publicador
A structured exercise that simulates several thinking archetypes, each emphasizing different assumptions and priorities, then synthesizes their perspectives. These are generated viewpoints from one system, not independent experts or independent evidence. Label factual claims, assumptions, and value judgments separately; verify consequential factual claims against external sources.
Use for a requested comparison of competing priorities or assumptions, including research planning decisions. The output is a decision aid: a conditional recommendation, unresolved questions, and a next step. It does not establish scientific validity by agreement or by the number of perspectives.
This skill runs within the current conversation. It has no bundled code, service API, credentials, or MindBook connection; use the host's available research tools when factual verification is needed. Do not claim that it creates hosted sessions, stores persistent memories, or contacts real experts.
The Council has three phases:
Based on the user's question, select 4–6 Council Members from the archetypes below unless the user requests another configuration. Choose lenses that examine different assumptions, constraints, or values. Agreement supported by the evidence is a valid outcome.
The 12 Archetypes:
Selection heuristic: Match the question type to the most productive tension:
These are starting points — adapt based on the specific question. The goal is to expose decision-relevant trade-offs and evidence gaps without a quota for disagreement or consensus.
Each Council Member delivers their perspective in this format:
🎭 [ARCHETYPE NAME]
Position: [One-sentence stance]
Reasoning: [2-4 sentences explaining their logic from their specific lens]
Basis: [Which given/verified facts, assumptions, or values support the position]
Key Risk They See: [The danger others might miss]
Decision Check: [Evidence or a changed constraint that would alter this position]
Critical rules for deliberation:
After all members speak, deliver:
⚖️ COUNCIL SYNTHESIS
Points of Convergence: [Shared conclusions, with their evidence and assumptions; agreement alone is not a confidence signal]
Core Tension: [The decision-relevant trade-off, or state that no substantive disagreement remains]
Open Question: [An overlooked issue or missing evidence; say if none was identified]
Recommended Path: [Conditional recommendation, next step, and what would change the choice]
Confidence Level: [Qualitative High / Medium / Low with reasons, tied to a specific claim; no percentage inferred from votes or rhetoric]
One Question to Sit With: [The question the user should keep thinking about after this session]
Before delivering the synthesis, check that every consequential factual claim traces to the evidence brief, each recommendation respects the constraints, and uncertainties are still visible. A critical unverified premise prevents a high-confidence recommendation. Distinguish confidence in the proposed next step from confidence in the eventual outcome. Record supported agreement and unresolved disagreements; do not resolve them by majority vote.
This is a qualitative review by the agent, not an automated validation gate or proof that the premises are true. Report unresolved checks explicitly.
The user can customize the Council:
See advanced configurations for domain mixes, optional editorial scoring, and bounded multi-round or synthesis-only output.
The Council works best on questions where:
The Council adds less value on:
If the question seems too simple for a full Council, say so — and offer a quick 2-perspective contrast instead.
User: "Quick council: should we run a pilot or begin the full experiment? We have ten weeks and 100 budget units. The pilot costs 20 units and takes two weeks; the full study costs 80 units and takes eight weeks. We don't know whether our assay is reproducible."
This is a fictional, manually worked illustration, not an experimentally validated outcome. The costs, durations, and uncertainty are given facts within the scenario. Assume sequential scheduling, no overlap, and that the pilot would assess the assay conditions needed for the study. Protecting interpretable results is a stated decision priority, not an empirical finding.
Synthesis: All three lenses support resolving the assay uncertainty; this agreement adds no independent evidence. The core tension is learning before commitment versus leaving no schedule or budget margin. The open question is whether the pilot's acceptance criteria can answer the reproducibility question under study conditions. Recommend defining those criteria and a stop rule before committing to either path. Confidence is medium in that next step and low in successful completion of the full study within the current limits. One question to resolve: "Which pilot result would actually change our decision?"
Sensitivity check: If the pilot costs 25 units, the sequential plan costs 105 and is infeasible without a scope or budget change. If relevant assay validation already exists, revisit whether a new pilot is needed. These changes should alter the recommendation even if every archetype initially agreed.
This specific prompt workflow and its optional scoring have not been benchmarked for decision accuracy. Multi-agent studies do not directly validate several voices generated in one conversation. Baltaji et al. (2024) report persona inconsistency and conformity in their tested collaborations. A September 2026 preprint by Ferreira et al. reports that debate did not reliably beat budget-matched sampling in its small-model setting. These bounded findings motivate checking evidence and assumptions; they do not establish a universal benefit or failure of council exercises.
Originally created by AHK Strategies, with inspiration from MindBook. The current MindBook FAQ, reviewed 2026-09-30, distinguishes its 12-archetype Council Chamber from a separate six-mind Mind Council. This repository's standalone exercise retains its own 3/4–6/6-perspective configurations; it is not a client for either hosted feature.
Structures a multi-perspective council exercise for decisions, research trade-offs, and creative challenges. Simulates thinking archetypes, separates evidence from assumptions and values, and synthesi
---name: consciousness-councildescription: Structures a multi-perspective council exercise for decisions, research trade-offs, and creative challenges…---# skills/consciousness-council/SKILL.mdResumen general
Structures a multi-perspective council exercise for decisions, research trade-offs, and creative challenges. Simulates thinking archetypes, separates evidence from assumptions and values, and synthesizes a conditional recommendation. Use when the user requests a council, panel, devil's advocate analysis, "mind council", or deliberate comparison of perspectives on a difficult choice.
Preguntas frecuentes
Descarga skills/consciousness-council/SKILL.md desde el publicador, o copia la ruta y abre el repositorio. PlugMCP lista la skill y no aloja el archivo.
El publicador lo mantiene en https://github.com/K-Dense-AI/scientific-agent-skills.
Claude, Cursor, Codex y Grok Build pueden leer un SKILL.md que añadas desde el repositorio del publicador. Grok Bot lo instala cuando envías en un chat el enlace a la carpeta de la skill en GitHub.
| "What hasn't been tried yet?" |
| Can chase novelty over reliability |