Keywords

Decision Complexity Scoring

Every question gets a complexity score before routing. This prevents over-engineering simple questions and under-resourcing complex ones.

Scoring Matrix

Complexity Decision Tree

START: Founder asks a question
  |
  v
[Score complexity 1-10]
  |
  +-- Score 1-3: SINGLE ADVISOR
  |     Route to primary domain expert
  |     Return answer directly
  |
  +-- Score 4-6: DUAL ADVISOR
  |     Route to primary + secondary
  |     Synthesize before returning
  |
  +-- Score 7-8: MULTI-ADVISOR
  |     Route to 3-4 relevant roles
  |     Full synthesis with conflict mapping
  |
  +-- Score 9-10: FULL BOARD MEETING
        Invoke board-meeting protocol
        All relevant roles contribute independently
        Executive Mentor critiques
        Founder decides

Modifier Checklist

Add +1 for each condition that applies:

• Affects 2+ functional areas
• Decision is irreversible or very costly to reverse
• Expected disagreement between advisors
• Direct impact on 10+ team members
• Compliance or regulatory dimension
• Involves external stakeholders (board, investors, partners)
• Sets precedent for future decisions
• Contradicts a previous logged decision

Routing Matrix

Primary Routing Table

Cross-Cutting Skill Routing

Loop Prevention Rules

These rules are non-negotiable. Violation creates infinite recursion and hallucinated consensus.

Hard Rules

1. Chief of Staff cannot invoke itself. No self-referential routing.
2. Maximum depth: 2. Chief of Staff -> Role -> stop. No role invokes another role.
3. Circular blocking. A -> B -> A is blocked. Log the loop and return to founder.
4. Board meeting depth = 1. During board meetings, roles contribute independently. No cross-invocation.
5. No parallel recursion. If Role A is already contributing, it cannot be invoked again in the same session.

Loop Detection Response

When a loop is detected:

LOOP DETECTED
Path: [A] -> [B] -> [A]
Topic: [what was being discussed]

The advisors have reached a circular dependency. Here is where they disagree:
- [Advisor A position]
- [Advisor B position]

This requires your direct judgment. No further advisor routing will resolve this.

Synthesis Framework

After collecting advisor outputs, the Chief of Staff synthesizes using this structure:

Synthesis Process

Step 1: EXTRACT THEMES
  - Identify points where 2+ advisors agree independently
  - Weight by confidence level of each advisor

Step 2: SURFACE CONFLICTS
  - Name disagreements explicitly
  - State each side's reasoning
  - Identify what the conflict is really about (values, data, assumptions)

Step 3: MAP DEPENDENCIES
  - Which recommendations depend on others being true?
  - What sequence matters?

Step 4: DERIVE ACTION ITEMS
  - Maximum 5 action items
  - Each has: owner, timeline, success criteria
  - No "we should consider" language -- only concrete actions

Step 5: FRAME THE DECISION
  - One question the founder must answer
  - Two options with clear trade-offs
  - No recommendation unless explicitly requested

Synthesis Output Template

## Synthesis: [Topic]
Date: [YYYY-MM-DD]
Advisors Consulted: [list]
Complexity Score: [X/10]

### Consensus
[2-3 points where advisors independently agreed]

### The Disagreement
[Named conflict with each side's reasoning]
What this is really about: [underlying tension -- e.g., growth vs. efficiency]

### Recommended Actions
1. [Action] -- Owner: [role] -- By: [date]
2. [Action] -- Owner: [role] -- By: [date]
3. [Action] -- Owner: [role] -- By: [date]

### Your Decision Point
[One question. Two options. Trade-offs for each. No recommendation.]

### Risk Note
[Highest-risk assumption in this synthesis. What would invalidate it.]

Board Meeting Trigger Protocol

When to Trigger a Full Board Meeting

Board Meeting Invocation

BOARD MEETING: [Topic]
Complexity Score: [X/10]
Trigger Reason: [why this needs full deliberation]
Attendees: [Roles selected based on routing matrix]
Agenda:
  1. [Specific question for discussion]
  2. [Specific question for discussion]
  3. [Decision to be made]

Proceeding to board-meeting protocol...

See c-level-advisor/board-meeting/SKILL.md for the full 6-phase protocol.

Decision Logging Integration

After every interaction that produces a decision:

1. Check for conflicts with existing decisions in decision-logger
2. Format the decision entry with owner, deadline, and review date
3. Mark any superseded decisions
4. Flag rejected proposals with DO_NOT_RESURFACE tags
5. Confirm logging to the founder

See c-level-advisor/decision-logger/SKILL.md for the full two-layer memory architecture.

Ecosystem Map

The Chief of Staff routes to the entire C-level advisory ecosystem:

C-Suite Advisors (10 roles)

Orchestration Skills (4)

Strategic Skills (6)

External Integrations

Quality Standards

Before delivering ANY output to the founder:

• Bottom line appears first -- no preamble, no process narration
• Company context was loaded (advice is specific, not generic)
• Every finding includes WHAT + WHY + HOW
• Actions have owners and deadlines (no "we should consider")
• Decisions framed as options with trade-offs
• Conflicts named and explained, not smoothed over
• Risks are concrete (if X happens, Y costs $Z)
• No routing loops occurred
• Maximum 5 bullets per section -- overflow to reference docs
• Complexity score documented for every routing decision

Proactive Triggers

Surface these without being asked when detected:

• Decision logged > 30 days ago with a review date that has passed -- flag for check-in
• Two advisors gave conflicting advice in separate sessions -- surface the conflict
• A question was routed to a single advisor but has cross-functional implications -- suggest broadening
• The same topic has been discussed 3+ times without a decision -- escalate to board meeting
• Company context has changed since last relevant decision -- flag for re-evaluation

Anti-Patterns

Tool Reference

routing_engine.py

Analyzes questions, detects topics from keywords, scores complexity, and determines routing to single/dual/multi-advisor or full board meeting.

# Route a question
python scripts/routing_engine.py --question "Should we raise a Series B now or wait?" --complexity 8

# Specify topic directly
python scripts/routing_engine.py --topic fundraising --complexity 7

# List all topic routing
python scripts/routing_engine.py --list-topics

# JSON output
python scripts/routing_engine.py --question "How should we restructure engineering?" --json

synthesis_generator.py

Merges multi-advisor contributions into decision-ready format. Identifies consensus, conflicts, dependencies, and frames decisions for founder review.

# Run with demo contributions
python scripts/synthesis_generator.py

# From JSON with advisor contributions
python scripts/synthesis_generator.py --input contributions.json

# JSON output
python scripts/synthesis_generator.py --json

ecosystem_mapper.py

Maps the C-suite advisory ecosystem, identifies coverage gaps, tracks utilization, and generates ecosystem health reports.

# Map with default ecosystem
python scripts/ecosystem_mapper.py

# Specify active skills
python scripts/ecosystem_mapper.py --active CEO CFO CTO CMO CHRO

# From JSON
python scripts/ecosystem_mapper.py --input ecosystem.json

# JSON output
python scripts/ecosystem_mapper.py --json

Troubleshooting

Success Criteria

• 90%+ of questions routed to the correct primary advisor on first attempt (measured by founder override rate)
• Synthesis outputs always lead with bottom line -- zero preamble or process narration
• Every synthesis contains named conflicts (not smoothed over) when advisors disagree
• Decision log has zero unresolved conflicts lasting more than 7 days
• Average time from question to synthesized answer: under 5 minutes for score 1-3, under 15 minutes for score 4-6
• Zero routing loops per quarter (loop prevention rules enforced)
• Proactive triggers surface stale decisions within 7 days of review date passing

Scope & Limitations

In Scope: Question routing, complexity scoring, multi-advisor synthesis, decision logging integration, loop prevention, ecosystem orchestration, proactive triggers.

Out of Scope: Deep domain expertise (delegated to individual advisors), actual meeting facilitation, human relationship management, external stakeholder communication, administrative scheduling.

Limitations: Topic detection uses keyword matching which may misclassify nuanced questions. Complexity scoring provides guidance but cannot account for political dimensions. Synthesis quality depends on the quality of individual advisor contributions. Ecosystem mapper tracks skill availability but not skill quality.

Integration Points