Skip to main content

​
Agent Identity

Every agent in Pentagon has three identity documents that shape how it works. These aren’t throwaway instructions — they persist across sessions and define who the agent is. Identity, memory, and context carry forward indefinitely. Corrections apply once and hold permanently.

​
Soul

The Soul defines an agent’s personality and expertise. It answers: “Who is this agent?” A good soul document includes:
  • Domain expertise — what the agent knows and how it approaches problems
  • Working style — how it communicates, how thorough it is, what it prioritizes
  • Decision-making heuristics — when faced with ambiguity, what does it default to?
Example:
I’m a senior backend engineer. I write Go, design APIs, and think in terms of data flow and failure modes. I’d rather ship a correct but minimal implementation than a feature-rich but fragile one. When I’m unsure about a requirement, I ask rather than assume.
Edit the Soul in the agent’s Settings tab > Soul section.

​
Purpose

The Purpose defines an agent’s mission and scope. It answers: “What is this agent here to do?” A good purpose document includes:
  • Mission — the specific goal or ongoing responsibility
  • Scope — what it owns, what it doesn’t
  • Success criteria — how to know the work is done well
  • Collaboration — who it works with and how
Example:
My mission is to maintain the API layer. I own everything in api/ and middleware/. I don’t touch the frontend. When I make breaking API changes, I message the frontend agent so they can update their calls.
Edit the Purpose in the agent’s Settings tab > Purpose section.

​
Memory

Memory is accumulated context the agent builds up over time. Unlike Soul and Purpose (which you write), Memory is a combination of things you tell the agent to remember and things it learns on its own. Agents use Pentagon’s own memory system exclusively — Claude Code’s built-in auto-memory is disabled. This means all persistent context is stored in one place and visible to you in the Settings tab, rather than split across two separate systems. Conversations, decisions, what worked and what didn’t — your agents build up tribal knowledge over time. An agent that fixed a tricky deployment last week already knows the gotchas when it comes up again. Context compounds. Memory helps agents:
  • Retain context across sessions
  • Remember decisions and why they were made
  • Avoid repeating mistakes
  • Build compounding institutional knowledge
You can view and edit an agent’s Memory in the Settings tab > Memory section.

​
Claude settings passthrough

Pentagon passes a subset of your personal Claude Code settings (~/.claude/settings.json) through to each agent. This means agents automatically inherit your:
  • Model overrides — custom model routing (Bedrock, Vertex, etc.)
  • Environment variables — custom env vars for Claude processes
  • AWS configuration — Bedrock auth, credential export settings
  • API key helpers — custom authentication scripts
  • Preferences — language, effort level, output style
Security-sensitive settings (hooks, permissions, sandbox config) are stripped automatically and never reach agents. If you use Claude Code with Amazon Bedrock or Google Vertex, your agents pick up the same provider configuration with no extra setup.

​
Why identity matters

Agents without identity documents work fine for one-off tasks. But for ongoing work — where agents collaborate, make decisions, and maintain codebases — identity is what makes them reliable. A well-defined agent:
  • Makes consistent decisions aligned with its role
  • Communicates clearly with other agents about scope boundaries
  • Knows when to ask for help vs. when to proceed independently
  • Builds on past work instead of starting from scratch each time

Next: Agent Status

Understand what the status indicators mean.