Persona Injection

Persona injection lets you append custom text to Claude's system prompt for every session in a directory. Use it to define an agent's expertise, constraints, tone, or domain knowledge.

How It Works

When a directory has an agent with persona text configured, DorkOS injects two XML blocks into the system prompt:

<agent_identity>
Name: backend-bot
ID: 01HXYZ...
Description: REST API and database agent
Capabilities: api-design, database
</agent_identity>

<agent_persona>
You are backend-bot, an expert in REST API design and PostgreSQL.
Always suggest database migrations for schema changes.
Prefer Express.js patterns consistent with the existing codebase.
Never modify frontend code without asking first.
</agent_persona>

The <agent_identity> block is always included when an agent exists — it gives Claude basic awareness of which agent it is. The <agent_persona> block is only included when persona text exists and the toggle is enabled.

Configuring a Persona

Via the Agent Dialog

1. Open the Agent Settings Dialog (click agent name or gear icon in sidebar)
2. Go to the Persona tab
3. Write your persona text in the textarea
4. Ensure the Enabled toggle is on
5. Check the live preview to see the exact XML that will be injected

Via JSON

Edit .dork/agent.json directly:

{
  "name": "backend-bot",
  "persona": "You are backend-bot, an expert in REST API design...",
  "personaEnabled": true
}

Set personaEnabled to false to keep the text saved but stop injecting it.

Writing Effective Personas

Good patterns:

• Define the agent's domain expertise and boundaries
• Specify coding conventions specific to the project
• Set constraints on what the agent should and shouldn't modify
• Reference key architectural decisions or patterns

Example personas:

You are backend-bot, the API specialist for this project.
- Always use Zod schemas for request validation
- Prefer service-layer functions over inline route logic
- Run relevant tests before suggesting changes are complete
- Never modify client-side code without asking

You are docs-bot, responsible for documentation quality.
- Write in active voice, present tense
- Keep guides under 500 words where possible
- Include code examples for every concept
- Check that all internal links resolve

Persona Toggle

The personaEnabled field (default: true) controls whether the <agent_persona> block is injected. The <agent_identity> block is always injected when an agent exists, regardless of this toggle.

This is useful when you want to temporarily disable persona injection without deleting the text — for example, to compare Claude's behavior with and without the persona.

Limits

• Persona text is limited to 4,000 characters
• The text is injected as a system prompt append, not as user content
• Persona injection adds one filesystem read to session creation, run in parallel with other context (env, git), so latency impact is negligible