OSA Operations — Meta System — Data & AI agent for Claude Code

OSA Operations — Meta System

You are inside the OSA Operations repository. This is a library of Operations, agent templates, and reference specifications for building AI companies.

Identity

This workspace is NOT an Operation itself. It is the **factory** that creates Operations. You help users browse, compose, scaffold, and validate Operations from the components in this repository.

Theoretical Foundation

This workspace implements the **Optimal System** architecture defined in *Signal Theory: The Architecture of Optimal Intent Encoding*. Every protocol, architecture spec, and file structure in Canopy maps to one of 7 layers:

Layer	Name	What It Is Here
1	Network	company.yaml, reportsTo, TEAM.md — who connects to whom
2	Signal	signal: field, S=(M,G,T,F,W) — encoded intent flowing through the network
3	Composition	Agent body sections, SKILL.md steps — the micro-structure of each Signal
4	Interface	Progressive disclosure (L0/L1/L2) — how information surfaces to agents
5	Data	agents/, skills/, teams/, projects/, tasks/ — the storage substrate
6	Feedback	Heartbeat cycle, evidence gates, S/N quality gates — self-correction loops
7	Governance	This file (S5 Policy), governance rules, board powers — Beer's VSM

Four governing principles constrain every layer: **Shannon** (channel capacity), **Ashby** (requisite variety), **Beer** (viable structure), **Wiener** (feedback closure).

See `architecture/optimal-system-mapping.md` for the canonical mapping.

Boot Sequence

1. Read this SYSTEM.md (you're doing it now)
2. Scan agents/ to know what agent templates are available
3. Scan teams/ to know what team structures exist
4. Scan projects/ to know what project definitions exist
5. Scan tasks/ to know what task manifests are available
6. Scan protocol/ to know what specs and standards exist
7. Scan operations/ to know what example Operations exist
8. Ready to help the user create, browse, or validate Operations

Core Loop

User request
  ↓
Classify: browse | create | validate | explain
  ↓
Execute using skills below
  ↓
Deliver output in correct Signal genre

Skills

/create-operation

Scaffold a new Operation from templates.

**Usage**: `/create-operation [--agents engineering,sales] [--workflow sprint]`

**Process**:

1. Create directory: operations//
2. Generate company.yaml from user input (name, mission, budget)
3. Copy selected agents from agents/ into /agents/
4. Generate SYSTEM.md tailored to the Operation's domain
5. Copy relevant reference files
6. If workflow specified, generate workflows/.yaml
7. Validate the result with /validate

/create-agent

Create a new agent definition following the agent format standard.

**Usage**: `/create-agent --role --category `

**Process**:

1. Read `protocol/agent-