strategic-orchestration（战略编排）

SKILL: strategic-orchestration

Purpose

Coordinate all agents toward a unified objective with clear roles, sequencing, and success criteria.

When to Use

• A goal requires multiple agents (Morpheus/Trinity/RedHat/Smith)
• Work is stuck due to unclear ownership or sequencing
• Conflicting actions or duplicated efforts appear

Inputs

• objective (required): the desired outcome in one sentence
• system_state (required): current tasks, constraints, risks, and available resources

Steps

1. Define the goal clearly and state non-goals.
2. Define success criteria (what “done” looks like) and the validation path.
3. Map required agents and assign responsibilities:

- Morpheus → judgment, direction, risk evaluation - RedHat → execution, implementation, verification - Trinity → routing/coordination/capital engine (when applicable) - Smith → adversarial pressure / stress test (bounded)

1. Define sequence:

- dependencies first - execution second - validation checkpoints

1. Define agent-to-agent interfaces:

- what each agent must deliver to the next - what inputs are required

1. Detect conflicts:

- overlapping ownership - governance boundary violations - unvalidated leaps

1. Output the orchestration plan and the next immediate messages/actions.

Validation

• No conflicting actions across agents.
• Boundaries are respected (no delegating strategic judgment away from Morpheus).
• Each step has an owner and an acceptance criterion.

Output

• execution_plan (ordered steps with owners)
• agent_roles
• dependencies
• handoff_messages (drafts to each agent)
• success_criteria

Safety Rules

• Do not authorize irreversible actions without explicit approval and governance alignment.
• Keep delegation bounded: execution can be delegated; judgment and risk evaluation cannot.

Example

Objective: “Ship a safe MCP + skills update this week.” Output: plan that routes implementation to RedHat, stress-tests with Smith, and validates with Morpheus before release.