Protocols Catalog
Protocols are non-negotiable rules that every agent loads at the start of every session. Unlike skills (reference knowledge) or workflows (step-by-step procedures), protocols define hard boundaries that agents cannot override or opt out of.
All Protocols
Section titled “All Protocols”| Name | Description | Required |
|---|---|---|
| memory | How to read, write, and maintain working memory between sessions | Yes (all) |
| scope-boundaries | What you own, what you never touch, workspace boundaries | Yes (all) |
| security | Security enforcement — hard stops and hard enforcers | Yes (all) |
| session-start | Ordered startup sequence — what to read and check every session | Yes (all) |
Protocol Details
Section titled “Protocol Details”memory
Section titled “memory”Defines how agents manage their working memory file (agent/Core/memory.md). Covers reading memory at session start, writing updates during work, and maintaining the file between sessions. This protocol ensures agents use version-controlled memory inside the project instead of Claude Code’s auto-memory system.
scope-boundaries
Section titled “scope-boundaries”Establishes what each agent owns and what it must never touch. Defines workspace boundaries so agents stay within their designated directories. Prevents agents from modifying files outside their scope — the scope-guard-files sensor enforces this automatically.
security
Section titled “security”Hard security rules: what agents must never do, what they must always check, and what requires human approval. Covers secrets handling, credential management, and security-critical operations. This is the foundation that the security-audit workflow builds on.
session-start
Section titled “session-start”The ordered startup sequence every agent follows at the beginning of each session. Specifies exactly which files to read and in what order: identity first, then memory, then protocols, then project status. Ensures consistent behavior across sessions.
How Protocols Differ from Other Abilities
Section titled “How Protocols Differ from Other Abilities”| Aspect | Protocols | Skills | Workflows |
|---|---|---|---|
| Loaded when | Every session, automatically | On demand, during specific work | When starting a multi-step activity |
| Required | Yes — cannot be removed | No — optional per agent | No — optional per agent |
| Purpose | Hard rules and boundaries | Domain knowledge and standards | Step-by-step procedures |
| Can be overridden | No | Agent adapts to context | Agent follows the steps |