Praetor Team Planning

Praetor should feel like a company operating system, not a multi-agent toy.

The chairman talks primarily to the CEO. The CEO decides whether the work needs

a mission, a planning team, a board briefing, execution, review, or escalation.

AI Agents are visible as company employees with roles, reporting lines,

responsibility, current work, and status. The user manages Agent accountability

through Praetor CEO, without micromanaging prompt/personality internals.

Core Flow

1. The chairman gives the CEO an idea, instruction, or question.

2. The CEO planner converts the instruction into explicit actions.

3. If the work needs structure, Praetor creates a mission.

4. The CEO forms a mission team with clear Agent owners, roles, and reporting lines.

5. The Project Manager opens role-specific delegations.

6. The team produces a board briefing for the chairman.

7. The chairman approves, revises, or authorizes execution.

8. Execution creates documents, decisions, open questions, run records, and

reviewer checks.

9. Closeout uses memory promotion before anything becomes durable company

knowledge.

New Project Discovery Flow

When the chairman says they want to start a new project and discuss the idea,

Praetor must treat the request as product discovery first, not immediate coding.

Expected flow:

1. The CEO opens a discovery conversation and asks only the missing high-impact

questions about purpose, target users, constraints, success criteria, and

non-goals.

2. While the idea is still fluid, the CEO may draft a project brief in the

conversation without creating a mission.

3. Once the shape is clear enough to preserve and coordinate, the CEO creates a

new_product / planning_only mission.

4. The CEO forms a planning team with clear Agent accountability. Typical

participants are CEO, Project Manager, Product Manager, Technical Architect

or Developer, Reviewer, File System Steward, and optional Legal, Security,

Design, Marketing, or Sales specialists.

5. The File System Steward creates visible project artifacts inside the bounded

workspace, such as Projects/<slug>/PROJECT_BRIEF.md,

Projects/<slug>/PRODUCT_SPEC.md, Projects/<slug>/TECH_SPEC.md,

Projects/<slug>/DECISIONS.md, Projects/<slug>/OPEN_QUESTIONS.md, and the

mission folder under Missions/<mission_id>/.

6. If multiple Agent perspectives matter, the CEO or PM convenes a Level 3

meeting with agenda, proposals, discussion, dissent, decisions, and action

items.

7. The CEO returns a Board Briefing with the recommended MVP, non-goals,

technical route, risks, open questions, and decisions needed.

8. Only after the chairman explicitly authorizes building, scaffolding, or

execution does the CEO convert the planning mission into an execution mission

or create a follow-up implementation mission.

The CEO must not treat brainstorms, raw meeting discussion, or tentative ideas

as durable company memory. Low-risk preferences, open questions, draft specs,

and mission-local notes may be saved automatically with evidence and rollback.

Board Briefing

BoardBriefing is the formal artifact between planning and execution.

It contains:

This keeps AI-to-AI discussion from becoming the product. Raw messages remain

evidence. The briefing is the owner-facing result.

Agent / Role Pattern

Default planning teams can include:

Praetor should create only the Agents needed for the mission. Agents have Roles;

mission-scoped Agents may later become durable company Agents when their

responsibility is recurring.

Escalation

The PM can decide low-risk sequencing and implementation details. The PM must

escalate:

The CEO can resolve ordinary business tradeoffs. The chairman decides anything

that changes authority, risk, budget, external commitments, or final direction.

Memory Boundary

Team planning creates structured records:

The CEO may automatically manage low-risk memory from these records according to

the memory-management policy. High-risk governance, strategy, legal, financial,

security, privacy, or external-commitment conclusions must remain drafts or

become owner decisions before they are durable policy.