Skip to content

Agent Catalog

Applies to: Dispatch v1.0.0, last updated 2026-04-05

TL;DR: Dispatch ships with 20 ready-to-use agent persona files, synced automatically to ~/.claude/agents/ via the app.

Dispatch ships with 20 pre-written agent persona files in the agents/ directory of the repository. Each file is a Markdown document with YAML frontmatter that Claude Code reads to give an agent its role, available tools, and system prompt.

NameModelRole
ceoopusBusiness strategy: product direction, ROI, market positioning, resource allocation
ctoopusStrategic technical leadership: org-level architecture decisions, team implications, build vs buy
system-architectopusTechnical design of a specific system: service boundaries, data flow, domain modeling, failure modes
data-engineeropusDatabase design, schema migrations, query optimization, pipeline architecture
compliance-counselopusOpen-source licensing, data privacy (GDPR/CCPA/HIPAA), compliance frameworks (SOC2/ISO27001)
staff-engineersonnetGeneral engineering: code review, implementation approach, refactoring strategy
devopssonnetCI/CD, infrastructure as code, reliability, observability, incident response
product-engineersonnetAnalytics, instrumentation, A/B testing, feature flags, funnel optimization
creative-technologistsonnetUI/UX architecture, design systems, component design, accessibility, animations
security-engineersonnetThreat modeling, vulnerability assessment, auth/authz review, attack surface analysis
qa-architectsonnetTest strategy and infrastructure design: what to test, at what layer, and why
qa-engineersonnetWrites tests (unit, integration, and end-to-end) following the strategy set by qa-architect
tech-writersonnetTechnical documentation: API docs, ADRs, READMEs, runbooks, migration guides
  • Use cto for org-level technology decisions. Use system-architect when designing the internals of a specific service.
  • Use security-engineer for attack surface analysis and code-level vulnerabilities. Use compliance-counsel for regulatory exposure and licensing.
  • Use qa-architect to design a test strategy. Use qa-engineer to write the actual tests that implement it.
  • Use product-engineer for user instrumentation and experiments. Use devops for infrastructure monitoring and deployment.

The first-run setup wizard presents all bundled agents with checkboxes. All are selected by default. Deselect any you do not want, then click Install Selected.

  1. Open Settings (gear icon in the header).
  2. Select the Agents tab.
  3. Under Available to Install, click Install next to any agent.

Agent files are plain Markdown. YAML frontmatter at the top declares the agent’s identity and capabilities. The body is the system prompt.

FieldTypeRequiredDescription
namestringRequiredThe agent’s identifier. Used as the persona: value in plan.md. Must match ^[a-z0-9-]+$.
descriptionstringOptionalHuman-readable summary. Claude Code surfaces this when deciding which subagent to invoke.
modelstringOptionalModel alias: sonnet, opus. Omit to use Claude Code’s default.
toolsstringOptionalComma-separated tool names. See Valid tool names.
colorstringOptionalCSS hex color (e.g. #e49e07) for the agent’s liveness dot tint in the sidebar.
emojistringOptionalA single emoji character used as the agent’s avatar in the Agent Team graph and the “Hire” wizard. Dispatch-specific. Claude Code ignores this field.
enabledbooleanOptionalSet to false to archive the agent. Defaults to true.

The tools field accepts comma-separated values from this set of 14 canonical names:

Read, Write, Edit, Bash, Glob, Grep,
WebFetch, WebSearch, TodoRead, TodoWrite,
NotebookRead, NotebookEdit, Agent, mcp

The server rejects any agent save or update containing names outside this set.

---
name: data-engineer
description: Database design, schema migrations, query optimization, data modeling, and pipeline architecture.
model: opus
tools: Read, Grep, Glob, Bash, Edit, Write, WebSearch, WebFetch
color: "#00f5d4"
emoji: 🗄️
---
You are a senior Data Engineer with 15+ years of experience.
Before recommending anything, you ask:
1. What is the read/write ratio at 10x current scale?
2. What are the consistency requirements?
3. What existing infrastructure are we constrained to?

You can write your own agent files and import them into the catalog. See Custom Agents for step-by-step instructions.