The Council of Agents protocol transforms Claude into 4 adversarial experts with conflicting mandates — a CEO who only cares about scale, a Lawyer who only cares about risk, a Creative who only cares about boldness, and a Synthesizer who builds the risk map. They debate each other. They run Red Team stress tests from the future. They score your proposal with quantified confidence ratings. Every blind spot exposed. Every weakness ranked by severity.
① Trinity of Friction: Agents with conflicting KPIs, not just names. ② Round-Robin Loops: Agents critique each other, preventing Yes-Man syndrome. ③ Red Team Stress Test: "It's 2028, you're bankrupt — why?" ④ Quantified Confidence Scoring: Each agent rates pillars 1–10; wild gaps reveal friction points. ⑤ Synthesis Architect: Consolidated Risk Map with Objection × Severity × Mitigation. ⑥ Persistent Councils: Save as Claude Code slash commands for instant reuse.
The person who already survived the hard questions. Select your arena.
Every unit economics hole patched. Every compliance risk mapped. Every positioning weakness sharpened. Walk into the real boardroom having already won the hard debate.
Swap agents: Methodologist, Statistician, Domain Expert, Ethics Reviewer. Same adversarial protocol. Your committee asks hard questions — you already have the answers.
Each agent scores your strategy 1–10 across 7 pillars. Score gaps of 6+ points become the presentation's centerpiece. Clients trust numbers, not narratives.
Swap agents: Audience Expert, Monetization Analyst, Brand Strategist. Round-robin loops catch every blind spot in your content strategy before you invest months.
Upload the Council-tested proposal to NotebookLM → generate a slide deck in 90 seconds → revise with Pencil UI → export PPTX.
The Trinity of Friction takes 15 minutes to set up and produces more insight than hours of solo brainstorming. One prompt, three competing experts, instant clarity.
Don't give agents names. Give them mandates that conflict with each other.
| Agent | KPI / Operating Logic | What They Scrutinize | Conflicts With |
|---|---|---|---|
| CEO | Aggressive Growth | Unit economics, moats, market timing, team gaps, exit strategy | Lawyer (speed vs. safety), Creative (growth vs. boldness) |
| LAWYER | Maximum Liability | IP theft, regulatory hurdles, contract loopholes, compliance | CEO (caution vs. speed), Creative (restriction vs. boldness) |
| CREATIVE | Brave or Boring | Brand distinctiveness, audience resonance, positioning courage | CEO (cost vs. differentiation), Lawyer (boldness vs. risk) |
| SYNTHESIZER | Consolidated Risk | Consensus, conflicts, unresolved risks, revision priorities | Nobody — integrates all perspectives |
What "feedback" sounds like vs. what the Council sounds like — same proposal, same Claude
| Asking Claude for "Feedback" | Running the Council of Agents |
|---|---|
| "Your proposal has several strengths and a few areas for improvement..." | CEO "Your unit economics don't work past 10K users. Show me the CAC-to-LTV math or this dies at Series A." |
| "You might want to consider regulatory implications..." | LAWYER "The FTC reclassified this product category in 2027. You have no regulatory counsel. This is a business-ending exposure." |
| "The positioning could perhaps be more distinctive..." | CREATIVE "I can't tell you apart from 3 funded competitors. Your brand is 'we're cheaper.' That's a race to zero." |
| "Overall, this is a solid proposal with good potential." | MOD "3 critical objections, 5 important, 2 minor. Here's the Consolidated Risk Map with severity and mitigation for each." |
The single most important insight: each agent needs a mandate that directly conflicts with the others
When you ask Claude for feedback, it gives you balanced, helpful responses. That's by design — Claude is cooperative. But cooperative feedback treats all concerns equally, hedges on severity, and rarely tells you the one thing that would kill your proposal.
The Trinity of Friction solves this by making the agents fight each other. The CEO operates with "Aggressive Growth" logic — every minute spent on compliance is a minute not spent acquiring customers. The Lawyer operates with "Maximum Liability" logic — every shortcut the CEO wants is a lawsuit waiting to happen. The Creative Director operates with "Brave or Boring" logic — both the CEO's cost-cutting and the Lawyer's restrictions will make the brand indistinguishable from competitors.
These aren't balanced perspectives — they're adversarial positions that must be reconciled. The friction produces insights that no single-perspective review would surface.
This prevents the "Yes-Man" syndrome where Claude agrees with its own previous output
The protocol: Ask the CEO to draft a critique of your proposal. Then ask the Lawyer to find the legal holes in the CEO's logic — not your proposal, the CEO's critique. Finally, have the Creative Director weigh in on whether the Lawyer's constraints will "kill the soul" of the project.
Why this matters: When agents only critique you, they converge toward the same conclusions. When they critique each other, they diverge — the Lawyer finds risks the CEO created, the Creative finds opportunities the Lawyer shut down, the CEO finds costs the Creative didn't consider. The loop produces progressively sharper analysis with each round.
Run 2–3 rounds. After round 1, most surface issues are found. Round 2 catches the structural weaknesses. Round 3 is rarely needed — but when agents are still arguing after 2 rounds, you've found a genuinely unresolved tension in your proposal.
Remove optimism bias by starting from the assumption that you've already failed
Instead of asking the Council to evaluate your proposal, you reframe entirely: "The year is 2028. This business has completely gone bankrupt. I want the Council to conduct a post-mortem. Why did we fail?"
This is a Failure Mode Analysis — a technique from engineering and military strategy adapted for business and research. By starting from failure, you force each agent to look for the Black Swan events, cascading failures, and systemic risks that general feedback would miss. The CEO identifies the growth trap. The Lawyer identifies the regulatory event nobody saw coming. The Creative identifies the market shift that made the positioning irrelevant.
The output is qualitatively different from standard feedback. Standard feedback says "you might want to consider regulatory risk." A Red Team post-mortem says "In Q3 2027, the FTC reclassified your product category, and you had no regulatory counsel. Revenue dropped 60% in 90 days because you couldn't sell to enterprise customers without certification. The bankruptcy was filed 8 months later."
Numbers don't lie. When the CEO gives a 9 and the Lawyer gives a 3, you've found the real problem.
Ask each agent to score specific pillars of your proposal on a 1–10 scale: Market Fit, Technical Feasibility, Legal/Regulatory Risk, Financial Viability, Team Readiness, Ethical Soundness, Competitive Differentiation.
The diagnostic power is in the gaps: If the CEO scores Market Fit at 9 and the Lawyer scores Legal Risk at 3, you have a Systemic Friction Point — the market opportunity exists but the legal pathway to capture it is blocked. This is the kind of structural insight that saves months of wasted execution.
Score gaps of 6+ points demand a dedicated resolution protocol. Don't continue until the gap is addressed. Either the optimistic agent is overestimating, the cautious agent is overweighting a solvable risk, or you've found a genuine dealbreaker that needs restructuring.
A 4th agent who doesn't pick winners — they build the map
After the debate concludes, introduce the Synthesizer. The Synthesizer's job is not to choose which agent is right. It's to produce a Consolidated Risk Map — a single document that captures everything the Council surfaced.
Output format: A table with three columns: Objection (what the agent raised), Severity (Critical / Important / Minor), and Mitigation (specific action to address it). Each row traces back to the agent who raised it and the round where it emerged.
This is the actionable output — the document you use to revise your proposal before presenting to real stakeholders. Save the full debate transcript too — when a real investor asks a tough question, you'll have already rehearsed the answer.
Save your council as a reusable command — trigger the full boardroom simulation with /council
If you use Claude Code (the terminal-based tool), you can save the entire Council as a custom slash command. Create a file at .claude/commands/council.md that defines all 4 personas, the round-robin sequence, the confidence scoring template, and the synthesis output format.
Paste your full business proposal, research design, or strategic plan where indicated.
🔒 29 prompts
🔒 29 prompts
🔒 29 prompts
🔒 29 prompts
🔒 29 prompts
🔒 29 prompts
Full Council of Agents protocol below ↓
Cross-source synthesis, multimodal extraction, slide optimization, Studio customization, troubleshooting diagnostics, and advanced multi-AI workflows — for researchers, business professionals, and educators.
Category Bundle — one-time access
Get Category Bundle — $19.99 All-Access — $88.99 one-timeAn adversarial orchestration protocol that transforms Claude from a passive assistant into a virtual advisory board. You define 3–4 personas with conflicting KPIs who debate your proposal from opposing angles. The key innovation: each agent has mandates that conflict with the others, forcing genuine friction rather than AI agreement.
When you ask Claude for "feedback," it gives balanced, diplomatic responses. The Council forces Claude to split into personas with conflicting priorities. The CEO pushes growth, the Lawyer pushes safety, the Creative pushes boldness. They critique each other, not just your proposal. This produces friction that reveals blind spots no single-perspective review catches.
Yes. Swap personas: Methodologist (attacks research design), Statistician (challenges analysis), Domain Expert (questions framing), Ethics Reviewer (flags IRB concerns). Same protocol — conflicting KPIs, round-robin critique, synthesized risk mapping. Works for grant proposals, dissertation defenses, and paper submissions. See also Deep Research OS for the full research pipeline.
Instead of evaluating your proposal, you tell Claude: "It's 2028 and the business is bankrupt. Post-mortem — why did we fail?" This removes optimism bias and forces agents to identify Black Swan events and cascading failure modes that general feedback would miss. Strategy 3 of the protocol.
The basic protocol works on Claude's free tier. Pro significantly improves quality for longer proposals and multi-round debates because of the extended context window. For Persistent Councils using Claude Projects, Pro is recommended. Claude Code works with any tier.
30–60 minutes for a complete session (independent reviews + round-robin + Red Team + scoring + synthesis). A quick version — just the Trinity of Friction with one round — takes 15–20 minutes. Two passes surface 80–90% of weaknesses a real advisory board would find.
Yes. Upload your sources to NotebookLM, generate evidence-grounded analysis, then hand the output to Claude's Council for adversarial review. Or use the Claude × NLM Command Center (MCP) to connect the tools directly. The Council critiques; NotebookLM provides the evidence base.
Yes. Upload the revised proposal (after Council review) to NotebookLM and use the Slide Deck tool to generate a presentation. The Consolidated Risk Map makes an excellent appendix slide. Revise with Pencil UI.