Skip to main content

What is a directive?

A directive is a bounded work instruction issued to the consigliere. It has a scope, an owner (a capo role), and optionally a requiresApproval flag. The consigliere reads your standing orders, picks the right capo, and executes within the rules you’ve set.

Step 1 — Get your context

Before issuing directives, pull the full operational context. This is the single call that returns everything:
Response shape:

Step 2 — Issue your first directive

The requiresApproval: true flag means the consigliere will queue the output for your review before sending anything. Start with approval on everything — remove it later for work you trust.

Step 3 — Check the approval queue

You’ll see the drafted email sequence ready for review. Approve it:
Or reject with feedback:

Step 4 — Read the Daily Report

After the consigliere executes, every move is logged. Your Daily Report surfaces the receipts:

Using BossMode in Claude Code

If you have the BossMode Claude Code skill installed:
This returns the same full context blob, formatted for Claude to consume directly. Issue a directive via Claude Code:

Using the MCP server

In any MCP-compatible runtime:

What happens under the hood

  1. The consigliere reads your standing orders to understand the rules
  2. The directive is routed to the appropriate capo role
  3. The capo executes within the engine scope (marketing, revenue, ops, etc.)
  4. If requiresApproval: true, output is queued for your review
  5. Every move is logged to the audit feed with a receipt
  6. The Daily Report surfaces outcomes the next morning

Next

Write standing orders

Standing orders are the rules the consigliere follows when the boss is away. Without them, autonomous moves stop at review. With them, delegated work stays bounded by policy.