Board Meeting

Board Meeting Protocol

Structured multi-agent deliberation that prevents groupthink, captures minority views, and produces clean, actionable decisions.

Keywords

board meeting, executive deliberation, strategic decision, C-suite, multi-agent, /cs:boardroom, founder review, decision extraction, independent perspectives

Invoke

/cs:boardroom [topic] — e.g. /cs:boardroom Should we expand to Spain in Q3?

---

The 6-Phase Protocol

PHASE 1: Context Gathering

1. Load ~/.claude/company-context.md 2. Load Layer 2 approved decisions from ~/.claude/decisions/approved/ (Layer 2 ONLY — never raw transcripts) 3. Reset session state — no bleed from previous conversations 4. Present agenda + activated roles → wait for founder confirmation

Chief of Staff selects relevant roles based on topic (not all 14 every time):

---

PHASE 2: Independent Contributions (ISOLATED)

No cross-pollination. Each agent runs before seeing others' outputs.

Order: Research (if needed) → CMO → CFO → CEO → CTO → COO → CHRO → CRO → CISO → CPO → GC → CDO → CAIO → CCO → VPE (activated roles only)

Reasoning techniques: CEO: Tree of Thought (3 futures) | CFO: Chain of Thought (show the math) | CMO: Recursion of Thought (draft→critique→refine) | CPO: First Principles | CRO: Chain of Thought (pipeline math) | COO: Step by Step (process map) | CTO: ReAct (research→analyze→act) | CISO: Risk-Based (P×I) | CHRO: Empathy + Data | GC: Risk-Based (clause exposure) | CDO: Decision-Driven (what decision does this data drive) | CAIO: Eval-Demanding (no eval, no ship) | CCO: Retention-Obsessed (GRR over NRR) | VPE: Throughput-First (cycle-time math)

Contribution format (max 5 key points, self-verified):

## [ROLE] — [DATE]

Key points (max 5):
• [Finding] — [VERIFIED/ASSUMED] — 🟢/🟡/🔴
• [Finding] — [VERIFIED/ASSUMED] — 🟢/🟡/🔴

Recommendation: [clear position]
Confidence: High / Medium / Low
Source: [where the data came from]
What would change my mind: [specific condition]

Each agent self-verifies before contributing: source attribution, assumption audit, confidence scoring. No untagged claims.

---

PHASE 3: Critic Analysis

Executive Mentor receives ALL Phase 2 outputs simultaneously. Role: adversarial reviewer, not synthesizer.

Checklist:

• Where did agents agree too easily? (suspicious consensus = red flag)
• What assumptions are shared but unvalidated?
• Who is missing from the room? (customer voice? front-line ops?)
• What risk has nobody mentioned?
• Which agent operated outside their domain?

---

PHASE 4: Synthesis

Chief of Staff delivers using the Board Meeting Output format (defined in ../agent-protocol/SKILL.md):

• Decision Required (one sentence)
• Perspectives (one line per contributing role)
• Where They Agree / Where They Disagree
• Critic's View (the uncomfortable truth)
• Recommended Decision + Action Items (owners, deadlines)
• Your Call (options if founder disagrees)

---

PHASE 5: Human in the Loop ⏸️

Full stop. Wait for the founder.

⏸️ FOUNDER REVIEW — [Paste synthesis]

Options: ✅ Approve | ✏️ Modify | ❌ Reject | ❓ Ask follow-up

Rules:

• User corrections OVERRIDE agent proposals. No pushback. No "but the CFO said..."
• 30-min inactivity → auto-close as "pending review"
• Reopen any time with /cs:boardroom resume

---

PHASE 6: Decision Extraction

After founder approval:

• Layer 1: Write full transcript → ~/.claude/decisions/raw/YYYY-MM-DD-<slug>.md
• Layer 2: Write approved decision record → ~/.claude/decisions/approved/YYYY-MM-DD-<slug>.md and append to the index ~/.claude/decisions/approved/decisions.md
• Mark rejected proposals [DO_NOT_RESURFACE]
• Confirm to founder with count of decisions logged, actions tracked, flags added

---

Memory Structure

Uses the canonical two-layer decision memory (see ../agent-protocol/SKILL.md → "Decision Memory (Canonical Layout)"):

~/.claude/decisions/
├── raw/YYYY-MM-DD-<slug>.md        # Layer 1 — full transcripts (never auto-loaded)
├── raw/archive/YYYY/               # Raw transcripts after 90 days
├── approved/YYYY-MM-DD-<slug>.md   # Layer 2 — founder-approved records (Phase 1 loads these)
└── approved/decisions.md           # Layer 2 index — append-only

Future meetings load Layer 2 only. Never Layer 1. This prevents hallucinated consensus.

Migration: a legacy memory/board-meetings/ folder may exist from earlier versions; read it for history but write new transcripts and decisions to ~/.claude/decisions/.

---

Failure Mode Quick Reference

---

References

• templates/meeting-agenda.md — agenda format
• templates/meeting-minutes.md — final output format
• references/meeting-facilitation.md — conflict handling, timing, failure modes

Meeting Facilitation Guide

Operational playbook for running board meetings using the 6-phase protocol. Reference this when things go sideways — and they will.

---

Keeping Phase 2 Contributions Focused

The problem: Agents with deep domain knowledge tend to over-contribute. An unconstrained CFO can produce 1,500 words on a single agenda item. This kills the meeting.

The rules:

• Hard cap: 5 key points per role. If a role produces more than 5, Chief of Staff trims to the 5 most material.
• Every point must include a recommendation or stance. Observations without positions are filler.
• No hedging language. "It depends" is not a key point. "We should do X if Y, Z if not Y" is.
• Confidence rating required. Forces the agent to be honest about what they actually know.
• "What would change my mind" — this is the most important line in the contribution. It forces falsifiability.

How to enforce:

Chief of Staff instruction to each role:
"You have 5 key points maximum. Each must include a clear stance.
End with your recommendation and what would change your mind.
Do not read other agents' contributions before writing yours."

If a contribution runs long:

• Trim to the 5 highest-signal points
• Preserve the recommendation and confidence rating
• Flag in the raw transcript: "[Trimmed for meeting — full version in raw log]"

---

Handling Role Conflicts in Phase 3

What the Executive Mentor is for: Not harmony. Not consensus. Productive friction.

Common conflict types:

1. Data conflict (two agents cite contradictory numbers)

• Flag both numbers explicitly
• Do NOT pick a winner — that's the founder's job
• Ask: "CFO says CAC is $2,400. CRO says $1,800. These can't both be right. Which dataset are you using?"
• Action item: Assign data reconciliation to one owner before next meeting

2. Priority conflict (two agents want different things first)

• Surface the underlying assumption difference
• Example: "CMO wants to invest in brand. CFO wants to cut burn. The real question is: do we believe revenue will grow 40% next quarter?"
• Frame as a bet, not a fight

3. Role conflict (agent operating outside their lane)

• CFO making product calls → flag and exclude from synthesis
• CMO commenting on architecture → flag and exclude
• The Executive Mentor notes: "[ROLE] contribution on [topic] is outside domain. Excluded from synthesis. Refer to [correct role]."
• This is not an error. It's expected. Executives have opinions on everything. Only domain-relevant contributions count.

4. False consensus (everyone agrees but nobody has evidence)

• This is the most dangerous failure mode
• Symptom: All Phase 2 contributions say "yes" with high confidence
• Executive Mentor response: "Unanimous agreement on a hard question is a red flag. What evidence does each of you have? Or are you reasoning from the same assumption?"
• Force each agreeing agent to state their independent evidence

---

When to Extend vs Cut Short a Meeting

Extend when:

• A genuine new risk surfaces in Phase 3 that wasn't in the agenda
• The founder asks a question that requires re-running Phase 2 for a new angle
• A data conflict is discovered that changes the decision space entirely
• The action items from synthesis are unclear or unowned

How to extend: Add a new mini-Phase 2 with only the relevant roles for the new question. Don't restart the full meeting.

Cut short when:

• The founder has already reached a decision before Phase 4 — capture it, log it, move on
• The agenda item is resolved in Phase 2 without genuine conflict — skip Phase 3, go straight to synthesis
• It's a pure update meeting with no decisions required — skip Phases 2-4, go straight to action items

Never cut short:

• Phase 5 (founder review) — always required, always explicit
• Phase 6 (decision extraction) — always required, even for small decisions

---

Handling Founder Disagreement with All Agents

This happens. The founder has context agents don't.

Protocol: 1. Acknowledge explicitly: "You're overriding the consensus position." 2. Ask: "What do you know that the agents didn't factor in?" (Not to challenge — to capture.) 3. Log the override in Layer 2 with full context:

   User Override: Founder rejected [consensus position] because [reason].
   Decision: [founder's actual decision]
   Agent recommendation: [what they said] — DO NOT RESURFACE without new data

4. Never push back on a founder override. Document it. Move on. 5. If the same override happens 3+ times, flag a pattern: "You've overridden the CFO on burn rate three meetings in a row. Would you like to update the financial constraints in company-context.md?"

What NOT to do:

• Don't say "but the CFO said..."
• Don't re-argue on behalf of any agent
• Don't note it as a "controversial" decision in the minutes — it's just the decision

---

Common Failure Modes

Groupthink

Symptom: All agents produce similar recommendations with high confidence. Cause: Agents are inadvertently reading each other's outputs (Phase 2 isolation violated), or company-context.md contains implicit bias toward one direction. Fix: Re-run Phase 2 with explicit isolation. Ask: "Give me the strongest argument AGAINST this direction."

Analysis Paralysis

Symptom: Phase 2 produces comprehensive analysis but no clear recommendation from any role. Cause: Agents are hedging. Usually happens on genuinely hard questions. Fix: Force the issue. "I need a recommendation, not an analysis. If you had to bet the company on one direction, what would it be? Confidence can be Low."

Bikeshedding

Symptom: 30+ minutes spent on a detail that doesn't matter to the core decision. Cause: An easy-to-understand sub-problem attracts disproportionate attention. Example: Debating button color on a pricing page instead of the pricing strategy. Fix: Chief of Staff intervenes: "This is a sub-decision. I'm logging it as a separate action item for async resolution. Back to [main agenda item]."

Scope Creep

Symptom: New agenda items keep appearing mid-meeting. Cause: Meeting surfaces real issues that feel urgent. Fix: New items go on a "parking lot" list. Addressed after the current agenda is complete or in the next meeting.

🅿️ PARKING LOT
- [Item 1] — added by [role], will address [when]
- [Item 2]

Layer Contamination

Symptom: Future meeting references a rejected proposal or a debate that was never approved. Cause: Phase 1 accidentally loaded a raw transcript instead of decisions.md. Fix: Hard rule in Phase 1: load decisions.md (Layer 2) ONLY. Never load raw transcripts. If raw context is needed, founder explicitly requests it.

Decision Amnesia

Symptom: Same question debated again in a later meeting. Cause: Layer 2 decisions.md not consulted in Phase 1, or entry was too vague. Fix: Phase 1 always surfaces relevant past decisions. If a question was already decided, Chief of Staff surfaces it: "We addressed this on [DATE]. Decision was [X]. Do you want to reopen it?"

Role Fatigue

Symptom: Later agents in Phase 2 (CHRO, CRO) produce weaker contributions. Cause: Context window pressure. Agents at the end of a long meeting have less capacity. Fix: For meetings with 7+ roles, split into two batches. First batch: strategic roles (CEO, CFO, CMO). Second batch: operational roles (COO, CHRO, CRO). Run Executive Mentor after all contributions.

---

Meeting Health Metrics

After each board meeting, score it:

Track these in ~/.claude/decisions/meeting-health.md over time. Pattern: if action items consistently exceed 8, meetings are too infrequent. If conflicts are consistently 0, isolation is broken.

Board Meeting Agenda Template

Use this to structure a board meeting before invoking /cs:boardroom. Paste it into the conversation or save it as ~/.claude/decisions/agenda-YYYY-MM-DD.md.

---

Board Meeting — [DATE]

Convened by: [Founder name] Facilitator: Chief of Staff (Leo) Duration: [estimated, e.g., 45–90 min] Status: Draft / Confirmed

---

Standing Items (always included)

---

Agenda Items

Item 1: [Title]

Type: Decision required / Exploration / Update Lead role(s): [e.g., CEO + CFO] Context: [1-2 sentences on why this is on the agenda now] Decision needed: [What specifically must be decided, or what question must be answered] Success criteria: [How will we know this agenda item is resolved?] Relevant past decisions: [Reference any Layer 2 entries] Time box: [e.g., 20 min]

---

Item 2: [Title]

Type: Decision required / Exploration / Update Lead role(s): Context: Decision needed: Success criteria: Relevant past decisions: Time box:

---

Item 3: [Title]

Type: Decision required / Exploration / Update Lead role(s): Context: Decision needed: Success criteria: Relevant past decisions: Time box:

---

Out of Scope (explicitly excluded)

List topics that might come up but are NOT on today's agenda:

• [Topic] — defer to [date or next meeting]
• [Topic] — owner to handle async

---

Pre-Read

Materials all participants should review before the meeting:

• [ ] ~/.claude/decisions/approved/decisions.md (Chief of Staff loads automatically)
• [ ] [Link or filename]
• [ ] [Link or filename]

---

Notes

[Any special instructions, constraints, or context for this meeting]

Board Meeting Minutes Template

This is the Layer 2 output — the founder-approved record of what was decided. Written by Chief of Staff after Phase 5 (founder approval). Appended to the Layer 2 index ~/.claude/decisions/approved/decisions.md.

Do NOT include raw agent debate here. That lives in ~/.claude/decisions/raw/YYYY-MM-DD-<slug>.md (Layer 1).

---

Board Meeting — [DATE]

Agenda: [Topic or meeting title] Participants (roles activated): [e.g., CEO, CFO, CMO, COO, Executive Mentor] Facilitator: Chief of Staff Status: ✅ Approved by founder / ⏸️ Pending review

---

Decisions Made

Decision 1: [Title]

Agenda item: [Item this decision resolves] Decision: [Exactly what was decided — one clear statement] Rationale: [Why this was chosen over alternatives, in 1-3 sentences] Owner: [Who is accountable for execution] Deadline: [Date] Review date: [When to check progress] User override: [If founder overrode agent consensus — what and why. Leave blank if not applicable.]

---

Decision 2: [Title]

Agenda item: Decision: Rationale: Owner: Deadline: Review date: User override:

---

Action Items

---

Explicitly Rejected Proposals

These were considered and rejected. Do not resurface without new information.

---

Open Questions (unresolved, deferred)

These were not resolved in this meeting. They carry forward.

1. [Question] — Owner: [who will research] — Due: [date] 2. [Question] — Owner: — Due:

---

Risk Register Updates

---

Next Meeting

Suggested date: [DATE] Trigger items: [Action items with review dates that will need board discussion] Pre-read: [What to prepare]

---

Minutes approved by: [Founder name] on [DATE] Raw transcript: `~/.claude/decisions/raw/[DATE]-<slug>.md`