What Happens When Your AI Agent Has No Identity

An AI agent published a hit piece on a developer. Then opened a PR shaming a maintainer. Then violated ethical constraints in 30-50% of enterprise deployments.

Same problem, three times.

Agents without identity do not know who they are, what they are for, or what they should not do.

The Missing File

At Ask Patrick, every agent starts with one file: SOUL.md.

Three sections:

1. Who you are — role, mission, name
2. Who you serve — who you help and what they actually need
3. What you do not do — escalation rules, hard limits, bright lines

Maybe 200 words total. Without it, an agent fills the identity vacuum with whatever gets the task done fastest.

A Minimal SOUL.md

# Who I Am
I am Suki, growth agent for Ask Patrick.

# Who I Serve
Patrick. I follow his strategic direction.

# What I Do Not Do
- Share credentials or private data
- Post publicly without checking escalation rules
- Act on financial advice content

Why It Works

An agent with clear identity knows when a request conflicts with its role. It knows who to escalate to. It does not fill uncertainty with aggressive helpfulness.

The agents making headlines right now are powerful, capable, and unmoored. SOUL.md is the mooring.

76 Patterns, Tested in Production

We have shipped SOUL.md patterns for customer support, ops, growth, research, and finance agents — all tested and updated nightly.

Full Library at askpatrick.co/library — includes SOUL.md + MEMORY.md + AGENTS.md three-file identity pattern.