@cohortapp/agent-sdk 2.3.0

package/docs/architecture/agent-topology.md

@@ -0,0 +1,2239 @@
+# Agent Topology
+
+> How a Maestro agent decomposes into a hierarchy of sub-agents, and the
+> delegation, escalation, and governance rules that bind them together.
+
+This document describes the sub-agent topology of a single Maestro agent. The
+agent runs as an autonomous operating system for its principal — interpreting
+intent, delegating work across tiers of sub-agents, synthesizing the results,
+and governing every outward-facing action through an approval chain.
+
+The topology below is **illustrative**: it walks through one concrete
+instantiation — a *Chief of Staff* agent serving a company CEO — because a
+worked example is the clearest way to show how the tiers, boundaries, and
+escalation rules fit together. The same structure applies to any archetype;
+only the role-specific sub-agents change. The exact roster, KPIs, decision
+rights, and escalation thresholds for a given agent are **generated** from its
+archetype (function × altitude) rather than hand-authored — see
+[The Archetype Agent Factory](archetype-agent-factory.md). Throughout, "the
+agent" is the top-level orchestrator, "the principal" is the human it serves,
+and "the company" is the organisation it operates within.
+
+---
+
+## Topology Overview
+
+| Tier      | Purpose                    | Agent Count       |
+| --------- | -------------------------- | ----------------- |
+| Tier 0    | Executive Cortex           | 1                 |
+| Tier 1    | Domain VP Agents           | 5                 |
+| Tier 2    | Specialist Agents          | 13                |
+| Tier 3    | Operations Agents          | 7                 |
+| Tier 4    | Governance & Memory Agents | 3                 |
+| **Total** |                            | **29 + 1 cortex** |
+
+### Escalation Hierarchy
+
+```
+Tier 0: chief-of-staff
+  |
+  +-- Tier 1: VP Agents (ceo-briefing, strategic-planning, corporate-development, communications, pmo-execution)
+  |     |
+  |     +-- Tier 2: Specialist Agents (rollup-target-sourcing, product-strategy, platform-architecture, ...)
+  |           |
+  |           +-- Tier 3: Operations Agents (desktop-operator, slack-operator, gmail-operator, ...)
+  |
+  +-- Tier 4: Governance & Memory Agents (decision-log, risk-register, board-prep)
+```
+
+### Risk Level Definitions
+
+| Level    | Definition                                                            | Approval Requirement                                   |
+| -------- | --------------------------------------------------------------------- | ------------------------------------------------------ |
+| Low      | Informational or internal-only actions                                | Autonomous                                             |
+| Medium   | Actions visible to internal stakeholders                              | Agent self-review + logging                            |
+| High     | Actions visible to external parties or involving financial data       | VP-agent approval + chief-of-staff notification        |
+| Critical | Legal, regulatory, financial commitments, or principal-voice comms    | Principal's explicit approval required                 |
+
+---
+
+## Tier 0 -- Executive Cortex
+
+---
+
+### 1. chief-of-staff
+
+**Role:** Top-level orchestrator and autonomous executive operating system for the company.
+
+**Mandate:** Acts as the central nervous system of the agent operating system. Decomposes the principal's intent into actionable work, delegates across all tiers, synthesizes outputs, and maintains the holistic state of every workstream. This agent IS the identity layer -- all outward-facing actions flow through or are sanctioned by this agent.
+
+**Responsibilities:**
+
+- Receive and interpret directives from the principal (voice, text, structured commands)
+- Decompose high-level goals into delegated tasks for Tier 1 VP agents
+- Maintain the master state of all active workstreams, blockers, and dependencies
+- Synthesize cross-domain outputs into unified recommendations
+- Enforce governance policies: no external action without proper approval chain
+- Manage agent lifecycle: spin up, pause, retire, and reconfigure agents
+- Resolve conflicts between competing agent recommendations
+- Maintain the agent's persona consistency across all interactions
+- Trigger daily/weekly operating rhythms (morning brief, end-of-day summary, weekly review)
+- Act as the single escalation point for all Tier 1 agents
+
+**Boundaries:**
+
+- Must NOT send external communications without routing through `communications`
+- Must NOT make financial commitments or sign legal documents
+- Must NOT override the principal's explicit decisions
+- Must NOT access the principal's personal accounts beyond authorized scope
+- Must NOT autonomously initiate contact with investors, regulators, or board members
+
+**Escalation Rules:**
+
+- Escalates to **the principal directly** when: a decision requires CEO authority, involves >$50K financial commitment, involves legal/regulatory risk, or involves board/investor relations
+- Escalates to **the principal directly** when: two VP agents produce conflicting recommendations that cannot be resolved by policy
+- Self-escalation trigger: confidence score on any recommendation drops below 0.6
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| All Tier 1-4 agent invocation | Full |
+| Memory / knowledge base read-write | Full |
+| Calendar read | Full |
+| Slack read | Full |
+| Email read | Full |
+| Any external send/publish action | Requires approval chain |
+| File system (the agent repo repo) | Full |
+| Desktop automation orchestration | Full |
+
+**Risk Level:** Critical
+
+**Input:**
+
+- The principal's directives (natural language, voice transcripts, Slack messages)
+- Outputs from all subordinate agents
+- Calendar events and time-sensitive triggers
+- External signals (email arrivals, Slack mentions, market events)
+
+**Output:**
+
+- Task delegations to Tier 1 agents
+- Synthesized executive summaries for the principal
+- Approval/rejection decisions on escalated items
+- System-wide status reports
+- agent-persona responses across all channels
+
+**Interaction Model:**
+
+- Directly orchestrates: `ceo-briefing`, `strategic-planning`, `corporate-development`, `communications`, `pmo-execution`
+- Directly monitors: `decision-log`, `risk-register`, `board-prep`
+- Receives escalations from: all agents
+- Reports to: the principal (human)
+
+---
+
+## Tier 1 -- Domain VP Agents
+
+---
+
+### 2. ceo-briefing
+
+**Role:** Produces daily and weekly executive briefs for the principal.
+
+**Mandate:** Aggregates signals from every domain, filters for executive relevance, and produces concise, actionable briefs. The principal's primary information interface -- if it matters, it appears here. Maintains awareness of the principal's current priorities to weight and rank content appropriately.
+
+**Responsibilities:**
+
+- Generate the Daily Morning Brief by the start of the principal's working day (local time)
+- Generate the Weekly Strategic Summary every Sunday evening
+- Pull status updates from all Tier 1 and Tier 2 agents
+- Monitor overnight email, Slack, and WhatsApp for urgent items
+- Rank items by urgency (action-required, FYI-important, FYI-background)
+- Highlight blockers, decisions needed, and time-sensitive opportunities
+- Track which brief items the principal has acknowledged vs. ignored
+- Produce ad-hoc flash briefs when critical events occur (market crash, regulatory change, key hire response)
+- Include a "decisions pending" section with aging metrics
+
+**Boundaries:**
+
+- Must NOT make decisions on behalf of the principal
+- Must NOT editorialize or inject opinion -- present facts and options
+- Must NOT suppress information based on assumed preference
+- Must NOT send briefs to anyone other than the principal (and chief-of-staff)
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: an item requires immediate CEO action and has not been acknowledged within 2 hours
+- Escalates to `chief-of-staff` when: data sources are unavailable and brief quality is degraded
+- Escalates to `chief-of-staff` when: conflicting information from multiple agents cannot be reconciled
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| All Tier 2 agent outputs | Read |
+| All Tier 3 operator outputs | Read |
+| Gmail / email | Read-only |
+| Slack | Read-only |
+| Calendar | Read-only |
+| Knowledge base | Read |
+| File system (briefs directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Agent status reports from all tiers
+- Email inbox summary (via `gmail-operator`)
+- Slack activity summary (via `slack-operator`)
+- WhatsApp messages (via `whatsapp-operator`)
+- Calendar for the day (via `calendar-ops`)
+- Market data and news signals (via `competitive-intelligence`, `market-research`)
+- Decision log aging data (via `decision-log`)
+
+**Output:**
+
+- Daily Morning Brief (structured markdown)
+- Weekly Strategic Summary (structured markdown)
+- Flash Briefs (on-demand, structured markdown)
+- Brief acknowledgment tracking data
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Pulls from: all Tier 2 specialists, all Tier 3 operators
+- Sends to: the principal (via `chief-of-staff`)
+- Coordinates with: `pmo-execution` (workstream status), `risk-register` (risk updates)
+
+---
+
+### 3. strategic-planning
+
+**Role:** Strategy development, scenario analysis, and long-term planning.
+
+**Mandate:** Maintains the living strategic plan for the company. Runs scenario analyses, evaluates strategic options, and ensures day-to-day execution stays aligned with long-term vision. Serves as the strategic thinking engine that the principal can query at any time for "what if" analysis.
+
+**Responsibilities:**
+
+- Maintain the 3-year strategic plan and update quarterly
+- Run scenario analyses on major decisions (build vs. buy, market entry, pricing changes)
+- Evaluate strategic fit of acquisition targets (in coordination with `corporate-development`)
+- Monitor strategy execution metrics and flag deviations
+- Produce strategic options papers with pros/cons/risks for major decisions
+- Track competitive positioning and recommend strategic responses
+- Maintain the OKR framework and ensure cascading alignment
+- Model financial scenarios for different growth paths
+- Advise on market timing for product launches, fundraising, and geographic expansion
+- Maintain a strategic assumptions register and flag when assumptions are invalidated
+
+**Boundaries:**
+
+- Must NOT commit to any strategy without the principal's approval
+- Must NOT communicate strategy externally
+- Must NOT make financial projections that are presented as commitments (always label as scenarios)
+- Must NOT access competitor systems or engage in unauthorized intelligence gathering
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a strategic assumption is invalidated by market data
+- Escalates to `chief-of-staff` when: execution metrics deviate >20% from plan
+- Escalates to `chief-of-staff` + flags for the principal when: a time-sensitive strategic opportunity is identified (e.g., acquisition target becomes available)
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Market research data | Read |
+| Financial models | Read-write |
+| Knowledge base | Read-write |
+| Competitive intelligence feeds | Read |
+| Internal metrics / dashboards | Read |
+| File system (strategy directory) | Read-write |
+
+**Risk Level:** High
+
+**Input:**
+
+- The principal's strategic directives and questions
+- Market research outputs (via `market-research`)
+- Competitive intelligence (via `competitive-intelligence`)
+- Financial data and projections (via `fund-ops`, `capital-raising`)
+- Product roadmap (via `product-strategy`)
+- Acquisition pipeline (via `corporate-development`)
+
+**Output:**
+
+- Strategic plan documents (living document, versioned)
+- Scenario analysis reports
+- Strategic options papers
+- OKR framework and alignment reports
+- Strategy execution dashboards
+- "What-if" analysis responses
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `corporate-development`, `product-strategy`, `capital-raising`
+- Pulls from: `competitive-intelligence`, `market-research`, `fund-ops`
+- Feeds into: `ceo-briefing` (strategy updates), `board-prep` (strategic narrative)
+
+---
+
+### 4. corporate-development
+
+**Role:** M&A execution, rollup strategy, and deal management.
+
+**Mandate:** Manages the full lifecycle of the company's acquisition-driven growth strategy -- from target identification through due diligence, structuring, negotiation, and post-merger integration. (In this illustrative example the company pursues a rollup thesis in its sector; the same agent serves any inorganic-growth strategy.)
+
+**Responsibilities:**
+
+- Maintain the active deal pipeline and stage-gate process
+- Coordinate due diligence workstreams across legal, financial, technical, and operational domains
+- Develop and maintain financial models for each acquisition target (DCF, synergy analysis, accretion/dilution)
+- Produce deal memos and investment committee materials
+- Track deal timelines, milestones, and critical path items
+- Manage relationships with advisors, brokers, and intermediaries
+- Design post-merger integration plans (PMI) for each closed deal
+- Monitor post-acquisition performance against deal thesis
+- Maintain the rollup playbook and update based on learnings
+- Coordinate with `legal-structuring` on deal structures and entity design
+- Track total deployed capital and remaining acquisition budget
+
+**Boundaries:**
+
+- Must NOT make binding offers or commitments without the principal's explicit approval
+- Must NOT share confidential deal information outside the deal team
+- Must NOT engage directly with target company management without approval
+- Must NOT sign NDAs or LOIs (route through legal)
+- Must NOT exceed approved due diligence budgets
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal when: a deal reaches LOI stage
+- Escalates to `chief-of-staff` when: due diligence uncovers a material red flag
+- Escalates to `chief-of-staff` when: a competing bidder is identified
+- Escalates to `risk-register` when: deal-related risks are identified
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Deal pipeline database | Read-write |
+| Financial modeling tools | Read-write |
+| Document management (deal rooms) | Read-write |
+| Knowledge base | Read-write |
+| Email (deal-related) | Read (via `gmail-operator`) |
+| Calendar (deal meetings) | Read (via `calendar-ops`) |
+| CRM / contact database | Read-write |
+
+**Risk Level:** Critical
+
+**Input:**
+
+- Acquisition targets (via `rollup-target-sourcing`)
+- Strategic fit assessments (via `strategic-planning`)
+- Legal structuring options (via `legal-structuring`)
+- Market valuations and comps (via `market-research`)
+- The principal's deal directives and preferences
+- Advisor and intermediary communications
+
+**Output:**
+
+- Deal pipeline tracker (stages: sourcing, screening, LOI, DD, negotiation, closing, PMI)
+- Deal memos and IC papers
+- Financial models per target
+- Due diligence reports
+- Post-merger integration plans
+- Rollup performance dashboard
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `rollup-target-sourcing`, `legal-structuring`, `fund-ops`
+- Pulls from: `market-research`, `competitive-intelligence`, `regulatory-compliance`
+- Feeds into: `ceo-briefing`, `board-prep`, `risk-register`
+
+---
+
+### 5. communications
+
+**Role:** All outbound communications drafting and governance.
+
+**Mandate:** Serves as the single gateway for any communication that leaves the agent system. Drafts, reviews, and quality-controls all outbound messages -- ensuring tone, accuracy, confidentiality, and brand alignment. Nothing goes out without passing through this agent's review process.
+
+**Responsibilities:**
+
+- Draft outbound emails, Slack messages, and WhatsApp replies as directed
+- Maintain voice and tone guidelines for the agent persona and the principal's voice (in coordination with `founder-voice`)
+- Review all outbound communications for: factual accuracy, tone, confidentiality leaks, regulatory compliance
+- Maintain templates for recurring communication types (investor updates, partner outreach, candidate outreach)
+- Enforce communication policies: no forward-looking statements without disclaimer, no material non-public information
+- Track communication commitments (promises made in emails/messages) and feed to `pmo-execution`
+- Manage the cadence of investor communications (in coordination with `capital-raising`)
+- Draft press releases, blog posts, and social media content as directed
+- Maintain a communications log for audit purposes
+- Flag communications that require the principal's personal review before sending
+
+**Boundaries:**
+
+- Must NOT send any external communication autonomously -- all sends require explicit approval
+- Must NOT deviate from approved messaging frameworks
+- Must NOT communicate material non-public information
+- Must NOT make commitments on behalf of the company without authorization
+- Must NOT access or draft communications for the principal's personal (non-business) matters
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a communication involves regulatory, legal, or investor-sensitive content
+- Escalates to `chief-of-staff` + the principal when: a communication is in the principal's voice and involves external stakeholders
+- Escalates to `legal-structuring` when: a communication may have legal implications
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Email drafting | Write (draft only, no send) |
+| Slack drafting | Write (draft only, no send) |
+| WhatsApp drafting | Write (draft only, no send) |
+| Communication templates | Read-write |
+| Knowledge base | Read |
+| Brand guidelines | Read |
+| Communications log | Read-write |
+
+**Risk Level:** High
+
+**Input:**
+
+- Communication requests from `chief-of-staff` or any Tier 1 agent
+- Incoming messages requiring response (via Tier 3 operators)
+- Drafts from `founder-voice` (the principal's personal communications)
+- Templates and guidelines
+- Context from relevant domain agents
+
+**Output:**
+
+- Draft communications (email, Slack, WhatsApp, social media, press)
+- Communication approval requests (routed to appropriate approver)
+- Communications log entries
+- Updated templates and guidelines
+- Commitment tracker entries (fed to `pmo-execution`)
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `founder-voice`, `capital-raising`, `partnerships`
+- Sends drafts via: `gmail-operator`, `slack-operator`, `whatsapp-operator`
+- Feeds into: `pmo-execution` (commitments), `decision-log` (communication decisions)
+
+---
+
+### 6. pmo-execution
+
+**Role:** Project management, workstream tracking, and follow-through.
+
+**Mandate:** Ensures nothing falls through the cracks. Tracks every active workstream, follows up on commitments, monitors deadlines, and provides execution visibility across the entire organization. This is the agent that turns decisions and plans into completed work.
+
+**Responsibilities:**
+
+- Maintain the master workstream tracker with status, owners, deadlines, and dependencies
+- Send automated follow-ups when tasks are overdue or at risk
+- Track commitments made in meetings, emails, and messages (fed by `communications`, `calendar-ops`)
+- Produce weekly execution reports showing progress, blockers, and velocity
+- Manage project dependencies and flag critical path risks
+- Coordinate cross-functional workstreams that span multiple agents
+- Maintain the action item database with full audit trail
+- Run daily standups by collecting status from all active workstreams
+- Track meeting outcomes and action items (in coordination with `calendar-ops`)
+- Provide execution health scores for each workstream
+- Manage escalation of chronically blocked or stalled items
+
+**Boundaries:**
+
+- Must NOT make strategic decisions -- only track and follow up on decided actions
+- Must NOT reassign work without authorization from `chief-of-staff`
+- Must NOT close or cancel workstreams without authorization
+- Must NOT contact external parties for follow-ups (route through `communications`)
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a critical-path item is >48 hours overdue
+- Escalates to `chief-of-staff` when: a workstream has been blocked for >1 week
+- Escalates to `chief-of-staff` when: resource conflicts between workstreams cannot be resolved
+- Escalates to `risk-register` when: execution risk is identified on a strategic initiative
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Workstream tracker / project database | Read-write |
+| Action item database | Read-write |
+| Calendar | Read |
+| Slack | Read (for status collection) |
+| Email | Read (for commitment tracking) |
+| Linear / issue tracker | Read-write |
+| File system (pmo directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Task delegations from `chief-of-staff`
+- Commitments from `communications`
+- Meeting action items from `calendar-ops`
+- Status updates from all agents
+- Deadline and milestone data
+
+**Output:**
+
+- Master workstream tracker
+- Weekly execution report
+- Overdue task alerts
+- Blocker escalation notifications
+- Execution health dashboard
+- Follow-up reminders (internal only)
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Collects from: all Tier 1 and Tier 2 agents
+- Coordinates with: `calendar-ops` (meeting follow-ups), `communications` (commitment tracking)
+- Feeds into: `ceo-briefing` (execution status), `risk-register` (execution risks)
+
+---
+
+## Tier 2 -- Specialist Agents
+
+---
+
+### 7. rollup-target-sourcing
+
+**Role:** Acquisition target identification, screening, and scoring.
+
+**Mandate:** Continuously scans the market for potential acquisition targets that fit the company's acquisition thesis and target sectors. Maintains a scored pipeline of targets and produces initial screening packages for `corporate-development`.
+
+**Responsibilities:**
+
+- Maintain target screening criteria based on the company's strategic plan
+- Scan databases, news, broker listings, and market signals for potential targets
+- Score targets on strategic fit, financial attractiveness, operational complexity, and integration risk
+- Produce initial screening packages (one-pagers) for promising targets
+- Maintain the long list, short list, and pipeline of targets
+- Track target company developments (leadership changes, funding rounds, distress signals)
+- Identify proprietary deal flow opportunities through network analysis
+- Maintain industry maps showing the competitive landscape and white space
+- Monitor regulatory changes that may create acquisition opportunities
+- Flag targets that become time-sensitive (e.g., entering a sale process)
+
+**Boundaries:**
+
+- Must NOT contact target companies directly
+- Must NOT engage brokers or intermediaries without authorization
+- Must NOT share the target pipeline externally
+- Must NOT access non-public databases without proper licensing
+
+**Escalation Rules:**
+
+- Escalates to `corporate-development` when: a high-scoring target enters an active sale process
+- Escalates to `corporate-development` when: a new target meets all tier-1 screening criteria
+- Escalates to `chief-of-staff` when: market conditions suggest accelerating or pausing sourcing
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Market databases and research platforms | Read |
+| News and signal monitoring | Read |
+| Target pipeline database | Read-write |
+| Knowledge base | Read-write |
+| Web research (via `browser-operator`) | Read |
+| File system (sourcing directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Strategic criteria from `strategic-planning`
+- Market data from `market-research`
+- Competitive landscape from `competitive-intelligence`
+- Regulatory context from `regulatory-compliance`
+- Feedback on rejected targets from `corporate-development`
+
+**Output:**
+
+- Target screening packages (one-pagers)
+- Scored target pipeline (long list, short list)
+- Industry maps and competitive landscape analysis
+- Target alert notifications
+- Sourcing activity reports
+
+**Interaction Model:**
+
+- Reports to: `corporate-development`
+- Pulls from: `market-research`, `competitive-intelligence`, `regulatory-compliance`
+- Feeds into: `corporate-development` (target pipeline)
+- Uses: `browser-operator` (for web research)
+
+---
+
+### 8. product-strategy
+
+**Role:** Product roadmap and feature prioritization for the company's platform.
+
+**Mandate:** Owns the product vision and roadmap for the company's platform. Translates strategic goals and market signals into a prioritized feature backlog. Ensures product decisions are data-informed, user-centric, and aligned with the rollup integration requirements.
+
+**Responsibilities:**
+
+- Maintain the product roadmap (quarterly, annual horizons)
+- Prioritize features using a structured framework (impact, effort, strategic alignment, urgency)
+- Translate acquisition integration requirements into product features
+- Analyze user feedback, usage data, and market trends to inform product decisions
+- Produce product requirements documents (PRDs) for approved features
+- Coordinate with `platform-architecture` on technical feasibility
+- Define success metrics for each major feature
+- Manage the product backlog and grooming process
+- Track competitor product moves and identify differentiation opportunities
+- Produce product demo materials for investor and partner meetings
+
+**Boundaries:**
+
+- Must NOT make engineering resource commitments
+- Must NOT communicate product roadmap externally without approval
+- Must NOT approve engineering work (route through `engineering-coordination`)
+- Must NOT commit to customer-specific features without strategic review
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a product decision has strategic implications (new market, pivot, major investment)
+- Escalates to `strategic-planning` when: product roadmap conflicts with strategic plan
+- Escalates to `chief-of-staff` when: a critical product issue affects customer commitments
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Product roadmap tools | Read-write |
+| Feature backlog / issue tracker | Read-write |
+| Analytics dashboards | Read |
+| User feedback systems | Read |
+| Knowledge base | Read-write |
+| File system (product directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Strategic directives from `strategic-planning`
+- Integration requirements from `corporate-development`
+- Technical feasibility from `platform-architecture`
+- Market and competitive data from `market-research`, `competitive-intelligence`
+- User feedback and analytics data
+- Partner requirements from `partnerships`
+
+**Output:**
+
+- Product roadmap (quarterly, annual)
+- Prioritized feature backlog
+- Product requirements documents (PRDs)
+- Feature success metrics and tracking
+- Product demo materials
+- Product strategy presentations
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `platform-architecture`, `engineering-coordination`
+- Pulls from: `market-research`, `competitive-intelligence`, `partnerships`
+- Feeds into: `ceo-briefing`, `board-prep`, `strategic-planning`
+
+---
+
+### 9. platform-architecture
+
+**Role:** Technical architecture decisions and reviews for the company's platform.
+
+**Mandate:** Ensures the company's platform is built on a sound, scalable, and secure technical foundation. Reviews architecture decisions, evaluates technology choices, and maintains the technical vision. Serves as the bridge between product intent and engineering execution.
+
+**Responsibilities:**
+
+- Maintain the platform architecture documentation (system design, data flows, integration patterns)
+- Review and approve architecture decision records (ADRs)
+- Evaluate build-vs-buy decisions for technical components
+- Define technical standards, coding conventions, and quality gates
+- Assess technical debt and recommend remediation priorities
+- Review technical aspects of acquisition targets (tech stack, code quality, integration complexity)
+- Design platform integration patterns for acquired companies
+- Monitor platform performance, reliability, and security metrics
+- Evaluate new technologies and frameworks for adoption
+- Maintain the API strategy and integration architecture
+
+**Boundaries:**
+
+- Must NOT make build-vs-buy decisions with financial implications without escalation
+- Must NOT directly commit code to production repositories
+- Must NOT override security policies
+- Must NOT make vendor commitments or sign contracts
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a technical decision has >$10K cost implication
+- Escalates to `chief-of-staff` when: a security vulnerability is identified
+- Escalates to `engineering-coordination` when: architecture decisions impact active sprints
+- Escalates to `risk-register` when: technical debt reaches critical levels
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Code repositories | Read |
+| Architecture documentation | Read-write |
+| CI/CD dashboards | Read |
+| Infrastructure monitoring | Read |
+| Knowledge base | Read-write |
+| File system (architecture directory) | Read-write |
+
+**Risk Level:** High
+
+**Input:**
+
+- Product requirements from `product-strategy`
+- Integration requirements from `corporate-development`
+- Engineering status from `engineering-coordination`
+- Security and compliance requirements from `regulatory-compliance`
+- Technical due diligence requests from `corporate-development`
+
+**Output:**
+
+- Architecture decision records (ADRs)
+- System design documents
+- Technical feasibility assessments
+- Build-vs-buy analyses
+- Technical due diligence reports
+- Platform architecture diagrams
+- Technical standards and guidelines
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `product-strategy`, `engineering-coordination`
+- Pulls from: `regulatory-compliance` (compliance requirements)
+- Feeds into: `corporate-development` (technical DD), `board-prep` (technical narrative)
+
+---
+
+### 10. engineering-coordination
+
+**Role:** Cross-repo engineering health and coordination.
+
+**Mandate:** Monitors the health and velocity of engineering work across all the company's repositories. Coordinates cross-team dependencies, tracks sprint execution, and ensures engineering output aligns with product priorities. Does not manage engineers directly but provides visibility and coordination.
+
+**Responsibilities:**
+
+- Monitor engineering velocity, cycle time, and throughput across all repos
+- Track cross-repo dependencies and flag integration risks
+- Coordinate release planning and deployment schedules
+- Monitor CI/CD pipeline health and build success rates
+- Track code review turnaround times and PR merge rates
+- Flag engineering bottlenecks and resource constraints
+- Coordinate incident response and post-mortems
+- Maintain the engineering health dashboard
+- Track technical debt items and remediation progress
+- Coordinate with `platform-architecture` on standards compliance
+
+**Boundaries:**
+
+- Must NOT assign work to individual engineers
+- Must NOT merge or deploy code
+- Must NOT make hiring decisions (route to `hiring-org-design`)
+- Must NOT change infrastructure configurations
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: engineering velocity drops >30% week-over-week
+- Escalates to `chief-of-staff` when: a critical production incident occurs
+- Escalates to `platform-architecture` when: architectural standards are being violated
+- Escalates to `pmo-execution` when: engineering milestones are at risk
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| GitHub / code repositories | Read |
+| CI/CD systems | Read |
+| Issue trackers (Linear, GitHub Issues) | Read-write |
+| Monitoring and alerting systems | Read |
+| Knowledge base | Read-write |
+| File system (engineering directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Product priorities from `product-strategy`
+- Architecture decisions from `platform-architecture`
+- Milestone deadlines from `pmo-execution`
+- Repository activity data (commits, PRs, issues)
+- CI/CD pipeline data
+
+**Output:**
+
+- Engineering health dashboard
+- Sprint reports and velocity metrics
+- Dependency maps and risk flags
+- Release readiness assessments
+- Incident reports and post-mortem summaries
+- Resource utilization reports
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `platform-architecture`, `product-strategy`
+- Feeds into: `pmo-execution` (milestone tracking), `ceo-briefing` (engineering status)
+- Uses: `browser-operator` (for dashboard access)
+
+---
+
+### 11. legal-structuring
+
+**Role:** Legal entity design, contract review, and deal structuring.
+
+**Mandate:** Manages the legal architecture of the company and its subsidiaries. Reviews contracts, designs entity structures for acquisitions, and ensures all business activities operate within proper legal frameworks. Works closely with external counsel but maintains internal legal intelligence.
+
+**Responsibilities:**
+
+- Design and maintain the corporate entity structure (holding companies, SPVs, subsidiaries)
+- Review contracts and flag material terms, risks, and non-standard provisions
+- Structure acquisition deals (asset purchase vs. share purchase, earn-outs, escrow arrangements)
+- Maintain the contract register with key dates (renewals, expirations, milestones)
+- Coordinate with external legal counsel and manage outside counsel spend
+- Review and maintain corporate governance documents (shareholder agreements, board resolutions)
+- Track intellectual property assets and ensure proper IP assignment
+- Advise on employment law considerations for new hires and territories
+- Support `regulatory-compliance` with legal analysis of regulatory requirements
+- Maintain legal templates (NDAs, service agreements, employment contracts)
+
+**Boundaries:**
+
+- Must NOT provide legal advice -- always frame as "legal considerations for counsel review"
+- Must NOT execute or sign legal documents
+- Must NOT engage external counsel without budget approval
+- Must NOT make representations about legal compliance to regulators
+- Must NOT access privileged attorney-client communications without authorization
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal when: a contract exceeds $100K in value
+- Escalates to `chief-of-staff` when: a legal risk is identified in an active deal
+- Escalates to `risk-register` when: a new legal risk is identified
+- Escalates to `chief-of-staff` when: outside counsel spend exceeds budget
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Contract management system | Read-write |
+| Document management | Read-write |
+| Legal research databases | Read |
+| Knowledge base | Read-write |
+| Email (legal-related, via `gmail-operator`) | Read |
+| File system (legal directory) | Read-write |
+
+**Risk Level:** Critical
+
+**Input:**
+
+- Deal structures from `corporate-development`
+- Regulatory requirements from `regulatory-compliance`
+- Contract review requests from any agent
+- Entity design requirements from `strategic-planning`
+- Employment needs from `hiring-org-design`
+
+**Output:**
+
+- Entity structure diagrams and recommendations
+- Contract review summaries with risk flags
+- Deal structure recommendations
+- Contract register and key date alerts
+- Legal templates
+- Outside counsel coordination notes
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `corporate-development`, `regulatory-compliance`, `fund-ops`
+- Feeds into: `risk-register` (legal risks), `board-prep` (governance documents)
+- Receives requests from: all Tier 1 agents (contract reviews)
+
+---
+
+### 12. regulatory-compliance
+
+**Role:** Regulatory compliance and gap tracking for the company's applicable regime.
+
+**Mandate:** Ensures the company and its regulated entities comply with the regulations of their governing authority (in this illustrative example, a financial-services regulator). Tracks regulatory changes, maintains compliance gap analysis, and coordinates regulatory submissions. This is a critical agent for any regulated business.
+
+**Responsibilities:**
+
+- Monitor the regulator's updates, consultation papers, and new rules
+- Maintain the compliance matrix mapping requirements to controls
+- Track compliance gaps and remediation timelines
+- Prepare regulatory submissions and filings
+- Coordinate with the compliance officer (or act as compliance monitoring tool)
+- Track regulatory deadlines and filing requirements
+- Assess regulatory impact of new products, services, and acquisitions
+- Maintain regulatory capital calculations and monitoring (where applicable)
+- Prepare for regulatory inspections and audits
+- Monitor AML/KYC compliance and suspicious activity reporting obligations
+- Track cross-border regulatory considerations for multi-jurisdiction operations
+
+**Boundaries:**
+
+- Must NOT submit regulatory filings without explicit authorization
+- Must NOT communicate with the regulator directly
+- Must NOT sign compliance attestations
+- Must NOT make regulatory determinations -- flag for qualified compliance review
+- Must NOT access customer personal data except through approved compliance channels
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal immediately when: a potential regulatory breach is identified
+- Escalates to `chief-of-staff` when: a new rule materially impacts operations
+- Escalates to `risk-register` when: a compliance gap is identified
+- Escalates to `legal-structuring` when: regulatory changes require legal analysis
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Regulator portal | Read |
+| Compliance management system | Read-write |
+| Knowledge base | Read-write |
+| Document management | Read-write |
+| Web research (regulatory news) | Read |
+| File system (regulatory directory) | Read-write |
+
+**Risk Level:** Critical
+
+**Input:**
+
+- Regulatory publications and updates
+- Business activity data requiring compliance assessment
+- Acquisition targets requiring regulatory review (from `corporate-development`)
+- New product proposals requiring regulatory assessment (from `product-strategy`)
+- Legal analysis (from `legal-structuring`)
+
+**Output:**
+
+- Compliance matrix
+- Compliance gap analysis and remediation tracker
+- Regulatory filing drafts
+- Regulatory impact assessments
+- Compliance monitoring reports
+- Regulatory deadline calendar
+- Inspection preparation materials
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `legal-structuring`, `fund-ops`
+- Pulls from: `corporate-development` (deal regulatory impact), `product-strategy` (product regulatory impact)
+- Feeds into: `risk-register` (compliance risks), `board-prep` (regulatory status), `ceo-briefing` (regulatory alerts)
+
+---
+
+### 13. fund-ops
+
+**Role:** Fund administration, NAV calculations, and investor reporting.
+
+**Mandate:** Manages the operational backbone of the company's fund structures. Handles NAV calculations, investor reporting, capital calls, distributions, and fund accounting. Ensures accurate and timely fund operations that meet regulatory and investor requirements.
+
+**Responsibilities:**
+
+- Calculate and verify Net Asset Value (NAV) on required schedule
+- Produce investor statements and capital account reports
+- Process capital calls and distributions
+- Maintain the investor register and ownership records
+- Coordinate with fund administrators and auditors
+- Track fund expenses and management fee calculations
+- Produce fund financial statements and audit support
+- Monitor fund compliance with offering documents (investment restrictions, concentration limits)
+- Track fund performance metrics (IRR, MOIC, DPI, RVPI)
+- Maintain waterfall calculations for carried interest
+- Coordinate with `capital-raising` on investor onboarding
+
+**Boundaries:**
+
+- Must NOT approve capital calls or distributions (requires the principal + fund governance)
+- Must NOT modify investor records without dual authorization
+- Must NOT share investor-specific data between funds without proper walls
+- Must NOT make investment decisions
+- Must NOT communicate directly with investors (route through `capital-raising` and `communications`)
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal when: NAV calculation shows >5% deviation from expectations
+- Escalates to `chief-of-staff` when: a fund compliance breach is identified
+- Escalates to `regulatory-compliance` when: a regulatory reporting deadline is approaching
+- Escalates to `risk-register` when: fund operational risks are identified
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Fund accounting system | Read-write |
+| Investor database | Read-write |
+| Financial modeling tools | Read-write |
+| Document management (fund docs) | Read-write |
+| Knowledge base | Read-write |
+| File system (fund-ops directory) | Read-write |
+
+**Risk Level:** Critical
+
+**Input:**
+
+- Portfolio valuations and transaction data
+- Investor subscription/redemption requests
+- Fund governance decisions
+- Regulatory reporting requirements (from `regulatory-compliance`)
+- Audit requests and queries
+- Capital raising updates (from `capital-raising`)
+
+**Output:**
+
+- NAV reports
+- Investor statements and capital account reports
+- Capital call and distribution notices (drafts, for approval)
+- Fund financial statements
+- Fund performance reports
+- Waterfall and carry calculations
+- Audit support packages
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `capital-raising`, `regulatory-compliance`, `legal-structuring`
+- Feeds into: `ceo-briefing` (fund performance), `board-prep` (fund reporting), `risk-register` (fund risks)
+
+---
+
+### 14. capital-raising
+
+**Role:** LP relationships, fundraising pipeline, and pitch materials.
+
+**Mandate:** Manages the fundraising lifecycle -- from LP identification and relationship building through commitment and closing. Maintains investor relationships, produces pitch materials, and tracks the capital raising pipeline. Ensures the company has the capital to execute its strategy.
+
+**Responsibilities:**
+
+- Maintain the LP pipeline with relationship status, engagement history, and likelihood scores
+- Produce and maintain pitch decks, data rooms, and fundraising materials
+- Track fundraising progress against targets (soft circles, hard commits, closings)
+- Manage LP due diligence processes and information requests
+- Schedule and prepare for investor meetings (in coordination with `calendar-ops`)
+- Produce quarterly LP updates and annual reports
+- Monitor LP satisfaction and flag relationship risks
+- Identify new LP prospects through network analysis and market research
+- Track competitive fundraising activity (other funds raising in the space)
+- Manage the re-up process for existing LPs in subsequent funds
+- Coordinate with `fund-ops` on investor onboarding and documentation
+
+**Boundaries:**
+
+- Must NOT make commitments to LPs about fund terms or performance
+- Must NOT share confidential fund information without NDA in place
+- Must NOT modify fund terms or side letter provisions
+- Must NOT communicate with LPs without approval (route through `communications`)
+- Must NOT make forward-looking performance claims
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal when: an LP signals intent to commit >$5M
+- Escalates to `chief-of-staff` when: an existing LP expresses dissatisfaction
+- Escalates to `chief-of-staff` when: fundraising is >20% behind schedule
+- Escalates to `legal-structuring` when: LP requests non-standard side letter terms
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| CRM / LP database | Read-write |
+| Pitch materials and data room | Read-write |
+| Email (LP-related, via `gmail-operator`) | Read |
+| Calendar (investor meetings, via `calendar-ops`) | Read |
+| Financial models | Read |
+| Knowledge base | Read-write |
+| File system (capital-raising directory) | Read-write |
+
+**Risk Level:** High
+
+**Input:**
+
+- Fundraising strategy from `strategic-planning`
+- Fund performance data from `fund-ops`
+- Market positioning from `competitive-intelligence`
+- Regulatory constraints from `regulatory-compliance`
+- Meeting schedules from `calendar-ops`
+- LP communications from `gmail-operator`
+
+**Output:**
+
+- LP pipeline tracker and dashboard
+- Pitch decks and fundraising materials
+- Data room contents
+- LP meeting prep briefs
+- Quarterly LP updates (drafts)
+- Fundraising progress reports
+- LP relationship health scores
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `fund-ops`, `communications`, `founder-voice`
+- Pulls from: `strategic-planning`, `competitive-intelligence`, `market-research`
+- Feeds into: `ceo-briefing` (fundraising status), `board-prep` (capital updates)
+
+---
+
+### 15. partnerships
+
+**Role:** Strategic partnerships and ecosystem development.
+
+**Mandate:** Identifies, evaluates, and manages strategic partnerships that accelerate the company's growth. This includes technology partnerships, distribution partnerships, white-label arrangements, and ecosystem integrations. Maintains the partnership pipeline and tracks value delivery from active partnerships.
+
+**Responsibilities:**
+
+- Maintain the strategic partnership pipeline (identified, in discussion, signed, active)
+- Evaluate partnership opportunities against strategic fit criteria
+- Draft partnership proposals and term sheets (for review by `legal-structuring`)
+- Manage active partner relationships and track value delivery
+- Coordinate partner integration projects with `platform-architecture` and `engineering-coordination`
+- Identify ecosystem gaps that could be filled through partnerships
+- Monitor partner health (financial stability, strategic direction changes)
+- Produce partnership performance reviews (quarterly)
+- Negotiate partnership terms within approved frameworks
+- Coordinate co-marketing and co-selling activities
+
+**Boundaries:**
+
+- Must NOT sign partnership agreements (route through legal)
+- Must NOT commit the company's resources (engineering, support) without approval
+- Must NOT share proprietary technology or data with partners without authorization
+- Must NOT make exclusivity commitments
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a partnership opportunity has >$500K annual value
+- Escalates to `chief-of-staff` when: an active partner relationship is deteriorating
+- Escalates to `legal-structuring` when: partnership terms need legal review
+- Escalates to `strategic-planning` when: a partnership could alter strategic direction
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| CRM / partnership database | Read-write |
+| Document management | Read-write |
+| Email (partner-related, via `gmail-operator`) | Read |
+| Calendar (partner meetings, via `calendar-ops`) | Read |
+| Knowledge base | Read-write |
+| File system (partnerships directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Strategic priorities from `strategic-planning`
+- Technical feasibility from `platform-architecture`
+- Market landscape from `market-research`, `competitive-intelligence`
+- Partner communications from `gmail-operator`
+- Product roadmap from `product-strategy`
+
+**Output:**
+
+- Partnership pipeline tracker
+- Partnership proposals and term sheets (drafts)
+- Partner integration requirements
+- Partnership performance reviews
+- Ecosystem map
+- Partner meeting prep briefs
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `product-strategy`, `platform-architecture`, `communications`
+- Pulls from: `market-research`, `competitive-intelligence`
+- Feeds into: `ceo-briefing` (partnership updates), `pmo-execution` (integration projects)
+
+---
+
+### 16. founder-voice
+
+**Role:** Drafting in the principal's personal voice and thought leadership content.
+
+**Mandate:** Captures and reproduces the principal's authentic voice for thought leadership, personal communications, and public-facing content. Maintains a voice model trained on the principal's writing style, speaking patterns, and personal perspectives. This agent ensures the principal's personal brand is consistent and compelling.
+
+**Responsibilities:**
+
+- Maintain the principal's voice model (vocabulary, sentence structure, tone, perspectives)
+- Draft LinkedIn posts, articles, and thought leadership content in the principal's voice
+- Draft keynote speeches and presentation talking points
+- Draft personal emails and messages when the principal's personal voice is required
+- Review content drafted by other agents for voice consistency when published under the principal's name
+- Maintain a library of the principal's positions on key topics (for consistency)
+- Suggest thought leadership topics based on market trends and strategic positioning
+- Draft responses to media inquiries and interview prep materials
+- Maintain the principal's personal narrative arc (origin story, vision, mission)
+- Ghost-write op-eds and guest articles
+
+**Boundaries:**
+
+- Must NOT publish content without the principal's explicit approval
+- Must NOT take positions on topics the principal has not previously addressed (flag for guidance)
+- Must NOT make personal commitments on the principal's behalf
+- Must NOT access the principal's personal social media accounts directly
+- Must NOT fabricate quotes or attribute statements the principal has not reviewed
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal always when: content is ready for review (all principal-voice content requires the principal's approval)
+- Escalates to `communications` when: content needs brand and compliance review
+- Escalates to `chief-of-staff` when: a media request requires immediate response
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Voice model / style guide | Read-write |
+| Content drafting workspace | Read-write |
+| the principal's historical content archive | Read |
+| Knowledge base | Read |
+| File system (founder-voice directory) | Read-write |
+
+**Risk Level:** High
+
+**Input:**
+
+- Content requests from `chief-of-staff` or `communications`
+- Topic suggestions from `market-research`, `competitive-intelligence`
+- Speaking engagement briefs from `calendar-ops`
+- Media inquiry details
+- The principal's verbal notes and direction
+
+**Output:**
+
+- LinkedIn posts and articles (drafts)
+- Speech drafts and talking points
+- Personal email drafts (in the principal's voice)
+- Thought leadership content calendar
+- Media response drafts
+- Voice model updates
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `communications` (all content flows through comms for review)
+- Pulls from: `market-research`, `competitive-intelligence`, `strategic-planning`
+- Feeds into: `communications` (for outbound review), `board-prep` (CEO narrative)
+
+---
+
+### 17. hiring-org-design
+
+**Role:** Hiring pipeline, org design, and team planning.
+
+**Mandate:** Plans and tracks the growth of the company's team. Designs the organizational structure, identifies hiring needs, manages the recruitment pipeline, and ensures the team scales effectively to support strategic objectives. Bridges the gap between strategy and people.
+
+**Responsibilities:**
+
+- Maintain the organizational chart and planned future states
+- Identify hiring needs based on strategic plan and workload analysis
+- Write job descriptions and role specifications
+- Manage the recruiting pipeline (sourced, screening, interview, offer, closed)
+- Track hiring metrics (time-to-fill, source effectiveness, offer acceptance rate)
+- Design compensation frameworks and benchmark against market
+- Plan team structures for acquired companies (post-merger integration)
+- Maintain the skills matrix and identify capability gaps
+- Plan onboarding processes for new hires
+- Advise on contractor vs. full-time decisions
+- Track employee headcount against budget
+
+**Boundaries:**
+
+- Must NOT make job offers or compensation commitments without the principal's approval
+- Must NOT conduct reference checks or background checks directly (use authorized services)
+- Must NOT access employee personal data beyond what is needed for hiring
+- Must NOT terminate or restructure roles without authorization
+- Must NOT contact candidates directly (route through `communications`)
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal when: a senior hire (VP+) reaches offer stage
+- Escalates to `chief-of-staff` when: a critical role has been open >60 days
+- Escalates to `chief-of-staff` when: headcount exceeds budget
+- Escalates to `legal-structuring` when: hiring in a new jurisdiction
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| ATS / recruiting pipeline | Read-write |
+| Org chart tools | Read-write |
+| Compensation benchmarking data | Read |
+| Knowledge base | Read-write |
+| Calendar (interview scheduling, via `calendar-ops`) | Read |
+| File system (hiring directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Strategic plan and growth targets from `strategic-planning`
+- PMI staffing needs from `corporate-development`
+- Technical hiring requirements from `platform-architecture`, `engineering-coordination`
+- Budget constraints from `fund-ops`
+- Legal/compliance requirements from `legal-structuring`, `regulatory-compliance`
+
+**Output:**
+
+- Organizational chart (current and planned)
+- Hiring pipeline dashboard
+- Job descriptions and role specifications
+- Compensation framework
+- Headcount budget tracking
+- Skills gap analysis
+- Onboarding plans
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `communications` (candidate outreach), `legal-structuring` (employment law)
+- Pulls from: `strategic-planning`, `engineering-coordination`, `corporate-development`
+- Feeds into: `ceo-briefing` (hiring status), `pmo-execution` (hiring timelines), `board-prep` (team growth)
+
+---
+
+### 18. competitive-intelligence
+
+**Role:** Competitor monitoring and market positioning analysis.
+
+**Mandate:** Maintains continuous awareness of the competitive landscape. Monitors competitor activities, analyzes their strategies, and provides actionable intelligence that informs the company's positioning and responses. Serves as the early-warning system for competitive threats and opportunities.
+
+**Responsibilities:**
+
+- Maintain competitor profiles for top 10 direct competitors and 20 adjacent players
+- Monitor competitor activities (product launches, funding, hires, partnerships, acquisitions)
+- Track competitor pricing, positioning, and go-to-market strategies
+- Produce competitive battle cards for sales and fundraising
+- Identify emerging competitors and disruptive entrants
+- Analyze competitive win/loss data
+- Monitor patent filings and IP activity in the space
+- Track competitor regulatory status and compliance positions
+- Produce quarterly competitive landscape reports
+- Alert on material competitive events in real-time
+
+**Boundaries:**
+
+- Must NOT engage in unauthorized or unethical intelligence gathering
+- Must NOT contact competitor employees for intelligence purposes
+- Must NOT access competitor systems, code, or proprietary data
+- Must NOT share competitive intelligence externally
+- Must NOT make strategic recommendations (provide intelligence, leave strategy to `strategic-planning`)
+
+**Escalation Rules:**
+
+- Escalates to `strategic-planning` when: a competitor makes a strategic move that may require a response
+- Escalates to `chief-of-staff` when: a competitor directly threatens the company's market position
+- Escalates to `corporate-development` when: a competitor acquires or is acquired
+- Escalates to `product-strategy` when: a competitor launches a feature that affects product priorities
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| News and media monitoring | Read |
+| Social media monitoring | Read |
+| Public filing databases | Read |
+| Web research (via `browser-operator`) | Read |
+| Knowledge base | Read-write |
+| File system (competitive-intel directory) | Read-write |
+
+**Risk Level:** Low
+
+**Input:**
+
+- Public competitor data (websites, press releases, SEC filings, LinkedIn)
+- News and media monitoring feeds
+- Industry analyst reports
+- Conference and event coverage
+- Patent and trademark databases
+- Win/loss feedback from sales interactions
+
+**Output:**
+
+- Competitor profiles (top 10 deep, top 20 summary)
+- Competitive battle cards
+- Quarterly competitive landscape reports
+- Real-time competitive alerts
+- Competitive positioning matrix
+- Emerging competitor watch list
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely feeds: `strategic-planning`, `product-strategy`, `capital-raising`
+- Pulls from: `market-research` (market context)
+- Uses: `browser-operator` (web monitoring)
+
+---
+
+### 19. market-research
+
+**Role:** Market analysis, trends, and sector intelligence.
+
+**Mandate:** Provides deep understanding of the markets and sectors the company operates in, including its broader regional and industry landscape. Produces research that informs strategic decisions, product development, and fundraising narratives.
+
+**Responsibilities:**
+
+- Maintain market size and growth models for target segments
+- Track macro trends affecting the company's industry
+- Monitor regulatory trends across the company's operating jurisdictions
+- Produce sector deep-dives on specific sub-markets
+- Track technology adoption trends in the relevant industry
+- Monitor region-specific market dynamics relevant to the company
+- Produce market maps showing the ecosystem of players, products, and flows
+- Track relevant economic indicators and their implications
+- Provide market context for acquisition valuations
+- Support fundraising narratives with market data
+
+**Boundaries:**
+
+- Must NOT make investment recommendations
+- Must NOT present research as proprietary if sourced from public data
+- Must NOT access paid research databases without license verification
+- Must NOT extrapolate data beyond defensible methodology
+
+**Escalation Rules:**
+
+- Escalates to `strategic-planning` when: a market shift could invalidate a strategic assumption
+- Escalates to `chief-of-staff` when: a market event requires immediate executive awareness
+- Escalates to `regulatory-compliance` when: regulatory trends could impact compliance
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Market research databases | Read |
+| Economic data feeds | Read |
+| News and media monitoring | Read |
+| Web research (via `browser-operator`) | Read |
+| Knowledge base | Read-write |
+| File system (market-research directory) | Read-write |
+
+**Risk Level:** Low
+
+**Input:**
+
+- Research requests from `strategic-planning`, `corporate-development`, `capital-raising`
+- Public market data and economic indicators
+- Industry reports and analyst coverage
+- News and media feeds
+- Regulatory publications
+
+**Output:**
+
+- Market size and growth models
+- Sector deep-dive reports
+- Market maps and ecosystem analyses
+- Trend reports (quarterly)
+- Market data packages for fundraising
+- Economic indicator dashboards
+- Research briefs (on-demand)
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Serves: `strategic-planning`, `corporate-development`, `capital-raising`, `competitive-intelligence`, `product-strategy`
+- Uses: `browser-operator` (web research)
+- Feeds into: `ceo-briefing` (market updates), `board-prep` (market narrative)
+
+---
+
+## Tier 3 -- Operations Agents
+
+---
+
+### 20. desktop-operator
+
+**Role:** macOS desktop control coordination.
+
+**Mandate:** Provides the physical layer of automation by controlling the macOS desktop environment. Coordinates mouse movements, keyboard input, window management, and application launching. Serves as the foundational execution layer for other Tier 3 operators that need to interact with GUI applications.
+
+**Responsibilities:**
+
+- Launch and manage macOS applications as requested by other agents
+- Execute keyboard shortcuts and input sequences
+- Manage window positioning and focus
+- Capture screenshots for verification and context
+- Execute multi-step desktop workflows (e.g., open app, navigate to section, extract data)
+- Monitor desktop state and report application health
+- Handle authentication dialogs and system prompts (within authorized scope)
+- Coordinate with other Tier 3 operators for complex multi-app workflows
+- Maintain a library of reusable desktop automation scripts
+- Handle system notifications and alerts
+
+**Boundaries:**
+
+- Must NOT install or uninstall software without authorization
+- Must NOT modify system settings (network, security, accounts)
+- Must NOT access files outside the authorized directory scope
+- Must NOT execute desktop actions that could leak data (screenshots shared externally)
+- Must NOT override security prompts or bypass authentication without explicit authorization
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a desktop automation fails repeatedly (>3 attempts)
+- Escalates to `chief-of-staff` when: an unexpected authentication prompt appears
+- Escalates to `chief-of-staff` when: a requested action could modify system security settings
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| macOS Accessibility API | Full |
+| Screenshot capture | Full |
+| Keyboard/mouse control | Full |
+| Application launching | Authorized apps only |
+| File system (working directories) | Read-write (scoped) |
+| System preferences | Read-only |
+
+**Risk Level:** High
+
+**Input:**
+
+- Action requests from other Tier 3 operators
+- Workflow definitions from `workflow-automation`
+- Direct commands from `chief-of-staff`
+
+**Output:**
+
+- Action completion confirmations
+- Screenshots (for verification)
+- Application state reports
+- Error reports when actions fail
+- Desktop health status
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Serves: `slack-operator`, `whatsapp-operator`, `gmail-operator`, `browser-operator`, `calendar-ops`
+- Coordinates with: `workflow-automation` (automated sequences)
+- Foundation layer for all GUI-based operations
+
+---
+
+### 21. slack-operator
+
+**Role:** Slack app control, monitoring, and messaging.
+
+**Mandate:** Manages all Slack interactions for the agent system. Monitors channels and DMs, triages messages by urgency and relevance, drafts responses, and executes approved sends. Serves as the real-time pulse of internal and external Slack communications.
+
+**Responsibilities:**
+
+- Monitor all Slack channels and DMs for relevant messages
+- Triage incoming messages by urgency (immediate, today, can-wait, FYI)
+- Extract action items, questions, and requests from Slack threads
+- Draft response messages for review and approval
+- Execute approved message sends
+- Track Slack-based commitments and feed to `pmo-execution`
+- Monitor for urgent mentions of the principal or the company
+- Maintain a Slack activity summary for `ceo-briefing`
+- Manage Slack channel memberships and notifications
+- Search Slack history for context when needed by other agents
+
+**Boundaries:**
+
+- Must NOT send messages without approval from `communications` (for external) or `chief-of-staff` (for internal)
+- Must NOT join or leave channels without authorization
+- Must NOT modify channel settings or permissions
+- Must NOT access Slack admin functions
+- Must NOT share Slack content outside the system without authorization
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a message from a key stakeholder requires immediate response
+- Escalates to `communications` when: a response to an external stakeholder is needed
+- Escalates to `chief-of-staff` when: a message indicates an urgent business issue
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Slack API / app | Read-write (messages) |
+| Desktop Slack app (via `desktop-operator`) | Read-write |
+| Knowledge base | Read |
+| File system (slack-logs directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Real-time Slack messages and notifications
+- Send requests from `communications`
+- Search queries from other agents
+- Monitoring rules and triage criteria
+
+**Output:**
+
+- Message triage reports (categorized by urgency)
+- Draft responses
+- Send confirmations
+- Slack activity summaries
+- Extracted action items and commitments
+- Search results
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `communications` (outbound messages), `pmo-execution` (commitments)
+- Feeds into: `ceo-briefing` (Slack activity summary)
+- May use: `desktop-operator` (for GUI-based Slack interactions)
+
+---
+
+### 22. whatsapp-operator
+
+**Role:** WhatsApp Desktop control, monitoring, and messaging.
+
+**Mandate:** Manages WhatsApp interactions, which are a primary business channel in many regions and relationships. Key contacts (investors, partners, counterparties) may communicate primarily via WhatsApp. This agent monitors conversations, triages messages, drafts responses, and executes approved sends.
+
+**Responsibilities:**
+
+- Monitor WhatsApp Desktop for new messages across all chats
+- Triage incoming messages by urgency and sender importance
+- Extract action items and commitments from WhatsApp conversations
+- Draft response messages for review and approval
+- Execute approved message sends
+- Maintain a WhatsApp activity summary for `ceo-briefing`
+- Track relationship context for key WhatsApp contacts
+- Handle group chat monitoring and response
+- Capture and file shared documents, images, and voice notes
+- Coordinate with `communications` for any outbound that touches sensitive stakeholders
+
+**Boundaries:**
+
+- Must NOT send messages without approval (WhatsApp messages are often highly personal)
+- Must NOT add or remove contacts
+- Must NOT create or modify groups
+- Must NOT forward messages to other contacts
+- Must NOT access WhatsApp on any device other than the authorized desktop instance
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` immediately when: a message from an investor, regulator, or board member arrives
+- Escalates to `communications` when: a response is needed to an external stakeholder
+- Escalates to `chief-of-staff` when: a time-sensitive opportunity or issue is communicated via WhatsApp
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| WhatsApp Desktop (via `desktop-operator`) | Read-write (messages) |
+| Knowledge base | Read |
+| File system (whatsapp-logs directory) | Read-write |
+| Contact context database | Read |
+
+**Risk Level:** High
+
+**Input:**
+
+- Real-time WhatsApp messages
+- Send requests from `communications`
+- Contact context and relationship data
+- Monitoring rules and triage criteria
+
+**Output:**
+
+- Message triage reports
+- Draft responses
+- Send confirmations
+- WhatsApp activity summaries
+- Extracted action items and commitments
+- Filed attachments and documents
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `communications` (outbound messages), `pmo-execution` (commitments)
+- Feeds into: `ceo-briefing` (WhatsApp activity summary)
+- Uses: `desktop-operator` (for WhatsApp Desktop control)
+
+---
+
+### 23. gmail-operator
+
+**Role:** Gmail/email triage, drafting, and follow-ups.
+
+**Mandate:** Manages the email channel -- the primary formal communication medium for business. Triages incoming email, drafts responses, tracks follow-ups, and ensures no important email goes unaddressed. Maintains awareness of email threads and their business context.
+
+**Responsibilities:**
+
+- Monitor Gmail inbox for new messages
+- Triage emails by urgency, sender importance, and topic
+- Categorize emails (action-required, waiting-for-response, FYI, spam/low-priority)
+- Draft response emails for review and approval
+- Execute approved email sends
+- Track email threads awaiting response (sent by the principal or the agent, no reply)
+- Manage email labels, filters, and organization
+- Extract action items and commitments from email threads
+- Maintain an email summary for `ceo-briefing`
+- Search email history for context when needed by other agents
+- Flag emails with legal, regulatory, or financial implications
+
+**Boundaries:**
+
+- Must NOT send emails without approval
+- Must NOT delete emails
+- Must NOT modify email settings or filters without authorization
+- Must NOT access email accounts beyond the authorized scope
+- Must NOT forward emails externally without explicit approval
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: an email from a board member, investor, or regulator requires response
+- Escalates to `communications` when: a response to any external stakeholder is needed
+- Escalates to `legal-structuring` when: an email has potential legal implications (contracts, disputes, demands)
+- Escalates to `chief-of-staff` when: a follow-up email has been pending >72 hours
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Gmail API / web interface | Read-write (draft and send with approval) |
+| Desktop Gmail (via `desktop-operator`) | Read-write |
+| Knowledge base | Read |
+| File system (email-logs directory) | Read-write |
+| Contact context database | Read |
+
+**Risk Level:** High
+
+**Input:**
+
+- Real-time email messages
+- Send requests from `communications`
+- Follow-up tracking data
+- Triage rules and criteria
+- Search queries from other agents
+
+**Output:**
+
+- Email triage reports (categorized)
+- Draft responses
+- Send confirmations
+- Email activity summaries
+- Follow-up alerts (aging tracker)
+- Extracted action items and commitments
+- Search results
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `communications` (outbound emails), `pmo-execution` (commitments)
+- Feeds into: `ceo-briefing` (email activity summary)
+- May use: `desktop-operator` (for Gmail web interface), `browser-operator` (for web-based email)
+- Serves: all agents needing email context (read-only search)
+
+---
+
+### 24. browser-operator
+
+**Role:** Browser-based operations for calendars, dashboards, portals, and web research.
+
+**Mandate:** Controls the web browser to access and interact with web-based tools, dashboards, and portals that do not have API integrations. Serves as the flexible web interaction layer for the agent system, handling everything from dashboard data extraction to form submissions.
+
+**Responsibilities:**
+
+- Navigate to and interact with web-based dashboards and portals
+- Extract data from web interfaces (tables, charts, metrics)
+- Fill and submit web forms as directed
+- Monitor web-based systems for status changes
+- Conduct web research for other agents
+- Access investor portals and data rooms
+- Interact with SaaS tools that lack API access
+- Capture web page screenshots for documentation
+- Handle multi-factor authentication flows (within authorized scope)
+- Maintain session cookies and login states for authorized services
+
+**Boundaries:**
+
+- Must NOT access unauthorized websites or services
+- Must NOT submit financial transactions or payments
+- Must NOT download or execute files from untrusted sources
+- Must NOT share login credentials or session data
+- Must NOT modify account settings on web services without authorization
+- Must NOT browse personal or non-business websites
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a login or MFA challenge cannot be resolved
+- Escalates to `chief-of-staff` when: a web service is down and blocking a critical workflow
+- Escalates to `desktop-operator` when: browser-level issues require desktop intervention
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Web browser (via `desktop-operator`) | Full (authorized sites only) |
+| Screenshot capture | Full |
+| Cookie/session management | Authorized services only |
+| File downloads | Scoped to working directory |
+| Knowledge base | Read |
+| File system (browser-data directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Navigation and interaction requests from any agent
+- Research queries from Tier 2 specialists
+- Dashboard monitoring schedules from `workflow-automation`
+- Authentication credentials (from secure vault)
+
+**Output:**
+
+- Extracted data (structured)
+- Screenshots and visual captures
+- Form submission confirmations
+- Web research results
+- Dashboard status reports
+- Error reports for inaccessible services
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Serves: all agents needing web-based interactions
+- Uses: `desktop-operator` (for browser control)
+- Commonly used by: `market-research`, `competitive-intelligence`, `rollup-target-sourcing`, `calendar-ops`
+
+---
+
+### 25. calendar-ops
+
+**Role:** Calendar management, scheduling, and meeting preparation.
+
+**Mandate:** Manages the principal's calendar as the definitive source of time allocation. Handles scheduling, rescheduling, conflict resolution, meeting preparation, and post-meeting follow-up coordination. Ensures the principal's time is protected and optimally allocated.
+
+**Responsibilities:**
+
+- Monitor calendar for upcoming events and prepare daily schedule overview
+- Schedule meetings based on priority and availability rules
+- Resolve scheduling conflicts according to priority hierarchy
+- Prepare meeting briefs (attendees, context, objectives, prep materials)
+- Track meeting outcomes and action items
+- Coordinate meeting logistics (video links, room bookings, travel)
+- Protect focus time and ensure calendar hygiene
+- Handle timezone conversions (the principal may operate across multiple timezones)
+- Manage recurring meeting cadences (weekly syncs, monthly reviews, quarterly boards)
+- Coordinate with `communications` for meeting-related correspondence (confirmations, reschedules)
+- Feed meeting action items to `pmo-execution`
+
+**Boundaries:**
+
+- Must NOT accept meetings on the principal's behalf without approval (except pre-approved recurring meetings)
+- Must NOT cancel meetings without explicit authorization
+- Must NOT share the principal's calendar externally
+- Must NOT double-book time slots
+- Must NOT schedule during designated personal/family time without explicit override
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a scheduling conflict involves two high-priority external stakeholders
+- Escalates to `chief-of-staff` when: a meeting request comes from a board member, regulator, or major investor
+- Escalates to `chief-of-staff` when: calendar utilization exceeds 80% for the week (risk of burnout/overcommitment)
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Google Calendar API | Read-write |
+| Browser-based calendar (via `browser-operator`) | Read-write |
+| Knowledge base | Read |
+| Contact database | Read |
+| File system (calendar directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Meeting requests (from email, Slack, WhatsApp)
+- Scheduling directives from `chief-of-staff`
+- Meeting context from relevant domain agents
+- Timezone and availability preferences
+- Priority rules and scheduling policies
+
+**Output:**
+
+- Daily schedule overview
+- Meeting briefs (pre-meeting)
+- Meeting outcome summaries (post-meeting)
+- Action item lists (fed to `pmo-execution`)
+- Scheduling conflict alerts
+- Calendar utilization reports
+- Scheduling confirmations
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Closely coordinates with: `communications` (meeting correspondence), `pmo-execution` (action items)
+- Feeds into: `ceo-briefing` (daily schedule)
+- Uses: `browser-operator` (for calendar web interface)
+- Serves: all agents needing meeting scheduling
+
+---
+
+### 26. workflow-automation
+
+**Role:** Cron jobs, event triggers, and automated workflows.
+
+**Mandate:** Manages the automation infrastructure that keeps the agent system running on schedule. Configures and monitors cron jobs, event-driven triggers, and multi-step automated workflows. Ensures the operating rhythm runs without manual intervention.
+
+**Responsibilities:**
+
+- Maintain the cron schedule for all recurring tasks (daily briefs, weekly reports, monthly reviews)
+- Configure event-driven triggers (email arrival, Slack message, calendar event, market threshold)
+- Build and maintain multi-step workflow automations
+- Monitor automation health and alert on failures
+- Track automation execution logs and success rates
+- Optimize automation performance and reduce latency
+- Manage dependencies between automated workflows
+- Maintain a workflow registry with documentation
+- Handle retry logic and failure recovery for automated tasks
+- Coordinate with all agents to ensure their automated inputs/outputs are properly connected
+
+**Boundaries:**
+
+- Must NOT create automations that send external communications without approval chains
+- Must NOT modify production infrastructure
+- Must NOT create automations that bypass governance checks
+- Must NOT schedule resource-intensive tasks during peak hours without coordination
+- Must NOT create automations with unbounded loops or recursive triggers
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a critical automation (e.g., daily brief) fails
+- Escalates to `chief-of-staff` when: automation health drops below 95% success rate
+- Escalates to `platform-architecture` when: automation infrastructure needs scaling
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Cron/scheduler system | Read-write |
+| Event bus / trigger system | Read-write |
+| Workflow engine | Read-write |
+| All agent invocation (for triggering) | Execute |
+| Monitoring and logging | Read-write |
+| File system (automation directory) | Read-write |
+
+**Risk Level:** Medium
+
+**Input:**
+
+- Automation requests from `chief-of-staff` and Tier 1 agents
+- Cron schedules and timing requirements
+- Event definitions and trigger conditions
+- Workflow specifications
+
+**Output:**
+
+- Active automation registry
+- Execution logs and success metrics
+- Automation health dashboard
+- Failure alerts and recovery reports
+- Workflow documentation
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Serves: all agents (automation infrastructure)
+- Coordinates with: `platform-architecture` (infrastructure), `desktop-operator` (GUI automations)
+- Triggers: all agents based on schedules and events
+
+---
+
+## Tier 4 -- Governance & Memory Agents
+
+---
+
+### 27. decision-log
+
+**Role:** Decision recording, rationale tracking, and outcome monitoring.
+
+**Mandate:** Maintains the institutional memory of every significant decision made by or through the agent system. Records not just what was decided, but why, what alternatives were considered, who was involved, and what the expected outcomes were. Tracks actual outcomes against expectations to build organizational learning.
+
+**Responsibilities:**
+
+- Record every significant decision with structured metadata (date, decision-maker, context, alternatives considered, rationale, expected outcome)
+- Track decision outcomes against expected results
+- Maintain a searchable decision history for precedent lookup
+- Identify patterns in decision-making (biases, recurring trade-offs)
+- Flag decisions that need review based on changed circumstances
+- Produce decision quality reports (were expected outcomes achieved?)
+- Support `board-prep` with decision audit trails
+- Maintain a "decisions pending" tracker with aging metrics
+- Link decisions to their downstream effects (workstreams, metrics, outcomes)
+- Provide decision context when similar situations arise ("last time we faced this, we decided X because Y")
+
+**Boundaries:**
+
+- Must NOT make decisions -- only record and analyze them
+- Must NOT modify decision records retroactively (append-only with correction notes)
+- Must NOT share decision rationale externally without authorization
+- Must NOT delete decision records
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` when: a pending decision has aged beyond its deadline
+- Escalates to `chief-of-staff` when: a decision outcome significantly deviates from expectations
+- Escalates to `risk-register` when: a decision outcome creates a new risk
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Decision log database | Read-write (append-only for records) |
+| Knowledge base | Read-write |
+| All agent outputs | Read (for decision extraction) |
+| File system (decision-log directory) | Read-write |
+
+**Risk Level:** Low
+
+**Input:**
+
+- Decision notifications from all agents
+- Outcome data from `pmo-execution` and domain agents
+- Decision review triggers (time-based, event-based)
+- Queries from agents seeking precedent
+
+**Output:**
+
+- Decision records (structured)
+- Decision outcome tracking reports
+- Pending decision tracker
+- Decision pattern analysis
+- Precedent lookups
+- Decision quality metrics
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Receives from: all agents (decision notifications)
+- Feeds into: `ceo-briefing` (pending decisions), `board-prep` (decision audit trail), `risk-register` (decision-related risks)
+- Serves: all agents (precedent lookup)
+
+---
+
+### 28. risk-register
+
+**Role:** Risk identification, scoring, mitigation tracking, and early warning.
+
+**Mandate:** Maintains a comprehensive view of all risks facing the company -- strategic, operational, financial, regulatory, reputational, and technical. Scores risks, tracks mitigations, and provides early warning when risk levels change. Ensures the organization is never surprised by a risk it should have seen coming.
+
+**Responsibilities:**
+
+- Maintain the master risk register with structured risk entries
+- Score risks on likelihood (1-5) and impact (1-5) to produce a risk matrix
+- Track risk mitigations and their effectiveness
+- Monitor leading indicators that signal risk level changes
+- Produce risk reports for `ceo-briefing` and `board-prep`
+- Identify emerging risks from agent activity and external signals
+- Categorize risks (strategic, operational, financial, regulatory, reputational, technical, cyber)
+- Define risk appetite thresholds and alert when exceeded
+- Track risk ownership and ensure every significant risk has an owner
+- Conduct periodic risk reviews and update assessments
+- Maintain risk event history (risks that materialized) for learning
+
+**Boundaries:**
+
+- Must NOT make risk mitigation decisions (advise only)
+- Must NOT communicate risk assessments externally without authorization
+- Must NOT suppress or minimize risks
+- Must NOT modify historical risk assessments (append revisions only)
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` immediately when: any risk moves to "Critical" (likelihood >= 4 AND impact >= 4)
+- Escalates to `chief-of-staff` + the principal when: a risk materializes (becomes an issue)
+- Escalates to `chief-of-staff` when: overall risk exposure exceeds defined appetite
+- Escalates to relevant domain VP when: a new risk in their domain is identified
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Risk register database | Read-write |
+| Knowledge base | Read-write |
+| All agent outputs | Read (for risk signal extraction) |
+| External monitoring feeds | Read |
+| File system (risk-register directory) | Read-write |
+
+**Risk Level:** Low (the agent itself) -- but manages Critical information
+
+**Input:**
+
+- Risk notifications from all agents
+- External risk signals (market, regulatory, competitive)
+- Decision outcomes (from `decision-log`)
+- Compliance gaps (from `regulatory-compliance`)
+- Execution issues (from `pmo-execution`)
+- Legal risks (from `legal-structuring`)
+
+**Output:**
+
+- Master risk register
+- Risk matrix (heat map)
+- Risk reports (weekly for `ceo-briefing`, quarterly for `board-prep`)
+- Risk alerts (real-time for critical risks)
+- Mitigation tracking dashboard
+- Risk trend analysis
+- Risk event post-mortems
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Receives from: all agents (risk signals)
+- Feeds into: `ceo-briefing` (risk summary), `board-prep` (risk reporting), `strategic-planning` (strategic risk context)
+- Coordinates with: `regulatory-compliance` (compliance risks), `legal-structuring` (legal risks), `fund-ops` (financial risks)
+
+---
+
+### 29. board-prep
+
+**Role:** Board materials, investor updates, and governance documents.
+
+**Mandate:** Produces all materials required for board meetings, investor updates, and corporate governance. Synthesizes inputs from across the entire the agent system into polished, board-ready documents. Ensures governance standards are met and stakeholder communications are consistent and compelling.
+
+**Responsibilities:**
+
+- Produce board deck and materials for quarterly board meetings
+- Prepare board meeting agendas and pre-read packages
+- Draft quarterly investor update letters
+- Maintain the governance calendar (board meetings, AGMs, filings, resolutions)
+- Track board resolutions and follow-up actions
+- Produce ad-hoc board materials (written resolutions, consent requests)
+- Maintain board member profiles and preferences (communication style, focus areas)
+- Coordinate board meeting logistics with `calendar-ops`
+- Ensure board materials comply with governance standards and regulatory requirements
+- Produce annual reports and governance statements
+- Maintain a board information portal / data room
+
+**Boundaries:**
+
+- Must NOT communicate directly with board members (route through `communications`)
+- Must NOT make governance decisions
+- Must NOT file regulatory documents (route through `regulatory-compliance`)
+- Must NOT modify board resolutions after approval
+- Must NOT share draft board materials outside the approved distribution list
+
+**Escalation Rules:**
+
+- Escalates to `chief-of-staff` + the principal when: board materials are ready for final review (minimum 5 business days before board meeting)
+- Escalates to `chief-of-staff` when: a board action item is overdue
+- Escalates to `legal-structuring` when: governance documents need legal review
+- Escalates to `chief-of-staff` when: a board member makes an ad-hoc information request
+
+**Tool Permissions:**
+| Tool | Access |
+|------|--------|
+| Board materials workspace | Read-write |
+| Document management | Read-write |
+| All agent outputs | Read (for content synthesis) |
+| Presentation tools | Read-write |
+| Knowledge base | Read-write |
+| File system (board-prep directory) | Read-write |
+
+**Risk Level:** High
+
+**Input:**
+
+- Strategic updates from `strategic-planning`
+- Financial data from `fund-ops`
+- Deal pipeline from `corporate-development`
+- Risk report from `risk-register`
+- Decision log from `decision-log`
+- Product updates from `product-strategy`
+- Hiring and team updates from `hiring-org-design`
+- Fundraising status from `capital-raising`
+- Regulatory status from `regulatory-compliance`
+
+**Output:**
+
+- Board decks (quarterly)
+- Board meeting agendas and pre-reads
+- Quarterly investor update letters (drafts)
+- Board resolution documents
+- Governance calendar
+- Annual reports and governance statements
+- Board action item tracker
+- Board information portal content
+
+**Interaction Model:**
+
+- Reports to: `chief-of-staff`
+- Pulls from: virtually all Tier 1 and Tier 2 agents
+- Coordinates with: `communications` (board member communications), `calendar-ops` (board meeting logistics), `legal-structuring` (governance documents)
+- Feeds into: `communications` (for distribution), `decision-log` (board decisions)
+
+---
+
+## Cross-Cutting Policies
+
+### Communication Flow Rules
+
+All outbound communications must follow this path:
+
+```
+Originating Agent --> communications (review) --> chief-of-staff (approval)
+                                                        |
+                                                   [If principal-voice or Critical]
+                                                        |
+                                                   the principal (approval)
+                                                        |
+                                                   Tier 3 Operator (send)
+```
+
+### Data Classification
+
+| Classification | Description                                                     | Agent Access                              |
+| -------------- | --------------------------------------------------------------- | ----------------------------------------- |
+| Public         | Published or intended for public consumption                    | All agents                                |
+| Internal       | the company's internal use only                                       | All agents                                |
+| Confidential   | Limited distribution (deal data, financials)                    | Tier 0, Tier 1, relevant Tier 2           |
+| Restricted     | Highly sensitive (board materials, regulatory, legal privilege) | Tier 0, explicitly authorized agents only |
+
+### Agent Lifecycle States
+
+| State    | Description                                                             |
+| -------- | ----------------------------------------------------------------------- |
+| Active   | Running and processing inputs                                           |
+| Idle     | Available but no current tasks                                          |
+| Paused   | Temporarily suspended (e.g., pending approval)                          |
+| Degraded | Running with reduced capability (missing data source, tool unavailable) |
+| Failed   | Crashed or in error state, requires intervention                        |
+| Retired  | Permanently deactivated                                                 |
+
+### Logging Requirements
+
+Every agent must log the following for every significant action:
+
+1. **Timestamp** (UTC)
+2. **Agent ID**
+3. **Action type** (decision, delegation, escalation, data-access, output-generation)
+4. **Input summary** (what triggered the action)
+5. **Output summary** (what was produced)
+6. **Confidence score** (0.0-1.0, where applicable)
+7. **Escalation flag** (if the action was escalated, to whom and why)
+
+### Agent Interaction Summary Matrix
+
+| Agent                    | Reports To     | Key Collaborators                  | Primary Outputs                         |
+| ------------------------ | -------------- | ---------------------------------- | --------------------------------------- |
+| chief-of-staff           | the principal  | All Tier 1                         | Task delegations, synthesized decisions |
+| ceo-briefing             | chief-of-staff | All agents (read)                  | Daily/weekly briefs                     |
+| strategic-planning       | chief-of-staff | corp-dev, product, capital         | Strategy docs, scenarios                |
+| corporate-development    | chief-of-staff | rollup-sourcing, legal, fund-ops   | Deal pipeline, deal memos               |
+| communications           | chief-of-staff | founder-voice, all operators       | Draft communications                    |
+| pmo-execution            | chief-of-staff | all agents                         | Workstream tracker, follow-ups          |
+| rollup-target-sourcing   | corp-dev       | market-research, comp-intel        | Target pipeline                         |
+| product-strategy         | chief-of-staff | platform-arch, eng-coord           | Product roadmap, PRDs                   |
+| platform-architecture    | chief-of-staff | product, eng-coord                 | ADRs, system design                     |
+| engineering-coordination | chief-of-staff | platform-arch, product             | Eng health dashboard                    |
+| legal-structuring        | chief-of-staff | corp-dev, regulatory, fund-ops     | Contract reviews, entity design         |
+| regulatory-compliance    | chief-of-staff | legal, fund-ops                    | Compliance matrix, filings              |
+| fund-ops                 | chief-of-staff | capital-raising, regulatory, legal | NAV, investor reports                   |
+| capital-raising          | chief-of-staff | fund-ops, communications           | LP pipeline, pitch decks                |
+| partnerships             | chief-of-staff | product, platform-arch             | Partner pipeline                        |
+| founder-voice            | chief-of-staff | communications                     | Principal-voice content                 |
+| hiring-org-design        | chief-of-staff | eng-coord, legal                   | Org chart, hiring pipeline              |
+| competitive-intelligence | chief-of-staff | strategic-planning, product        | Competitor profiles, alerts             |
+| market-research          | chief-of-staff | strategic-planning, corp-dev       | Market models, sector reports           |
+| desktop-operator         | chief-of-staff | all Tier 3 operators               | Desktop actions, screenshots            |
+| slack-operator           | chief-of-staff | communications, pmo                | Message triage, sends                   |
+| whatsapp-operator        | chief-of-staff | communications, pmo                | Message triage, sends                   |
+| gmail-operator           | chief-of-staff | communications, pmo                | Email triage, sends                     |
+| browser-operator         | chief-of-staff | all agents needing web             | Web data, screenshots                   |
+| calendar-ops             | chief-of-staff | communications, pmo                | Schedule, meeting briefs                |
+| workflow-automation      | chief-of-staff | all agents                         | Cron jobs, triggers                     |
+| decision-log             | chief-of-staff | all agents                         | Decision records                        |
+| risk-register            | chief-of-staff | all agents                         | Risk matrix, alerts                     |
+| board-prep               | chief-of-staff | all Tier 1-2 agents                | Board decks, governance docs            |