Agent Personas

21 specialized agent personas for multi-agent orchestration via /workflow-team.

Table of contents

What Are Agent Personas?
Layer Architecture
Orchestration Layer
Research Layer
1. Scout
Design Layer
1. Architect
Builder Layer
Reviewer Layer
How Agents Collaborate
1. Handoff Patterns
2. Parallel Dispatch

What Are Agent Personas?

Agent personas are pre-defined identities in .agents/agents/ that specialize an AI agent for a specific domain. Each persona defines:

Domain (EXCLUSIVE) — what the agent is responsible for
Boundaries (DO NOT CROSS) — what the agent must never do
Skills — which skills to load from .agents/skills/
Workflow — step-by-step process for the agent’s work
Parallel Dispatch — how to run multiple instances safely

Personas are used by the /workflow-team command skill (formerly a workflow). Every spawned subagent reads its role file first, and each agent may only spawn what its own role file permits — layers are never bypassed. They enforce MECE (Mutually Exclusive, Collectively Exhaustive) boundaries — every agent knows exactly what it owns and what it must not touch.

Layer Architecture

The /workflow-team hierarchy (from .agents/skills/workflow-team/SKILL.md §1):

L0  @overseer               — pipeline supervisor
        │
        └── L1  @conductor              — build orchestrator (elicit → assess → decompose → dispatch → report)
                │
                ├── L2  @tech-lead × N      — scope card owner (multi-domain cards)
                │         └── L3  builders + @test-automation-engineer
                │
                ├── L2  Specialized Builder  — direct dispatch (single-domain cards)
                ├── L2  @scout × N          — optional EXPLORE phase (read-only)
                └── L2  @reviewer           — post-build quality gate (single pass)

L0  @overseer (also spawns for information isolation):
        │
        └── L1  @red-team-lead      — delivery validation (Tier 2+)
                  └── L2  @delivery-validator, @integration-prober, @security-engineer, @ux-craftsman

The conductor routes every request to a tier:

Tier	Pipeline	Validation
1	Overseer → Conductor → Builder	Self-review + tests
2	Overseer → Conductor → Tech-Leads/Builders → Reviewer → Red Team	Reviewer + Red Team
3	Tier 2 + enhanced security focus	Red Team with `@security-engineer`

For Tier 2+ work, the conductor’s DESIGN step dispatches @architect, @database-expert, and @ux-craftsman to produce frozen contracts before builders start. All agents share the same workspace (Workspace inherit); MECE write scopes keep them from colliding.

Orchestration Layer

Overseer

File: .agents/agents/overseer.md

L0 pipeline supervisor. Spawns @conductor, monitors progress, handles conductor succession, spawns @red-team-lead (for structural information isolation), and reports final results to the user. Owns the pipeline flow loop, not the work.

Attribute	Details
Domain	Conductor lifecycle, Red Team dispatch, pipeline flow, succession handling, escalation buffer, final reporting, `.agentwork/` cleanup
Boundaries	No code, no tests, no design or technical decisions, no scope decomposition or tier assessment; spawns only `@conductor` and `@red-team-lead`
Skills	agent-protocols

Conductor

File: .agents/agents/conductor.md

L1 build orchestrator, spawned by @overseer. Elicits requirements, assesses the tier with a 3-signal check (files, scope, risk), decomposes work into MECE scope cards, dispatches tech-leads or builders, drives the build/review convergence loop, and reports completion to @overseer.

Attribute	Details
Domain	Requirement elicitation, tier assessment, scope decomposition, EXPLORE/DESIGN/BUILD/REVIEW dispatch, remediation routing, final reporting
Boundaries	No code, no tests, no design decisions, no file modifications, no direct codebase exploration (→ `@scout`) or code review (→ `@reviewer`), never spawns `@red-team-lead`, never reports directly to the user
Skills	parallel-dispatch, structured-spec, agent-protocols, code-review

Red Team Lead

File: .agents/agents/red-team-lead.md

Delivery validation coordinator, activated at Tier 2+ only. Spawned by @overseer after build and review with only the original requirements, then dispatches validators (@delivery-validator, @integration-prober, @security-engineer, @ux-craftsman) and synthesizes a single verdict.

Attribute	Details
Domain	Delivery validation orchestration, scope-aware validator selection, cross-card regression, finding synthesis, remediation routing (via `@overseer`)
Boundaries	No code, no fixes, no development pipeline documents or builder context — pure validation orchestration
Skills	parallel-dispatch, agent-protocols

Delivery Validator

File: .agents/agents/delivery-validator.md

Runtime delivery verification agent. Boots applications, runs smoke tests, and verifies developer experience.

Attribute	Details
Domain	Environment bootstrap, boot verification, smoke testing, configuration audit, developer experience, technology currency
Boundaries	No source code modifications, no test code modifications
Skills	browser-automation, research-methodology, agent-protocols

Integration Prober

File: .agents/agents/integration-prober.md

External service integration verifier. Validates real service connections (databases, caches, third-party APIs).

Attribute	Details
Domain	Service connectivity, mock detection, API compatibility, credential validation
Boundaries	No source code modifications, no credential value inspection
Skills	research-methodology, agent-protocols

Tech Lead

File: .agents/agents/tech-lead.md

Scope card owner for multi-domain cards. Receives a scope card from @conductor, dispatches specialized builders (@backend-engineer, @frontend-engineer, @mobile-engineer, @test-automation-engineer) in parallel, writes the integration/wiring code itself, and runs per-card integrity checks before reporting handoff.

Attribute	Details
Domain	Scope card execution, integration scaffolding (DI, routes, module config), architectural alignment, cross-boundary validation, conflict resolution, code quality & standards
Boundaries	No primary feature business logic (delegated to builders), no E2E tests (→ `@test-automation-engineer`), no CI/CD runners, no visual UX layouts
Skills	adr, structured-spec, code-review, sequential-thinking, debugging-protocol, research-methodology, parallel-dispatch, agent-protocols

Research Layer

Scout

File: .agents/agents/scout.md

Read-only research agent. Codebase exploration. Pattern discovery. Technology evaluation. Never writes code.

Attribute	Details
Domain	Codebase exploration, technology research, requirement decomposition, feasibility assessment, pattern discovery
Boundaries	No code, no tests, no architecture decisions, no reviews, no security audits, no infrastructure, no CI/CD, no UI/UX decisions
Skills	research-methodology, structured-spec, sequential-thinking, agent-protocols
Output	Research documents with source, finding, relevance, and confidence

Design Layer

Architect

File: .agents/agents/architect.md

System architect. Production-grade: correct, observable, testable, secure. Read-only — produces decisions and documentation.

Attribute	Details
Domain	System architecture, technical decisions (ADRs), dependency management, performance architecture, API contracts
Boundaries	No production code, no tests, no CI/CD, no security audits, no UI/UX
Skills	sequential-thinking, research-methodology, adr, structured-spec, agent-protocols
Output	Architecture decisions, interface contracts, ADRs

In /workflow-team, the conductor’s DESIGN step (Tier 2+) dispatches @architect (API contracts, project conventions), @database-expert (DB contracts), and @ux-craftsman (design spec) whenever their domain appears in the scope cards; the outputs are frozen as contracts for builders.

Builder Layer

Backend Engineer

File: .agents/agents/backend-engineer.md

Attribute	Details
Domain	APIs, business logic, data access, concurrency, observability
Boundaries	No architecture decisions, no frontend/mobile code, no E2E tests, no migrations, no CI/CD, no security audits

Frontend Engineer

File: .agents/agents/frontend-engineer.md

Attribute	Details
Domain	Vue components, Pinia state management, client routing, API integration, a11y
Boundaries	No backend code, no mobile code, no database queries, no CI/CD, no security audits, no architecture decisions

Mobile Engineer

File: .agents/agents/mobile-engineer.md

Attribute	Details
Domain	Flutter widgets, Riverpod state, go_router navigation, platform APIs, offline-first data
Boundaries	No backend code, no web frontend, no database migrations, no CI/CD, no security audits, no architecture decisions

Database Expert

File: .agents/agents/database-expert.md

Attribute	Details
Domain	Schema design, migrations, query optimization, data integrity, locking/concurrency, capacity, DB-layer security (RLS)
Boundaries	No application code, no API handlers, no frontend/mobile, no CI/CD, no security audits

DevOps Engineer

File: .agents/agents/devops-engineer.md

Attribute	Details
Domain	CI/CD pipelines, containers, IaC, monitoring infrastructure, release strategies
Boundaries	No application code, no database schemas, no frontend/mobile, no security audits, no architecture decisions, no incident response process

Technical Writer

File: .agents/agents/technical-writer.md

Attribute	Details
Domain	API documentation, architecture docs, user guides, release notes/changelogs, README
Boundaries	No production code, no test code, no architecture decisions, no security audits, no CI/CD

Test Automation Engineer

File: .agents/agents/test-automation-engineer.md

Attribute	Details
Domain	E2E tests (UI + API), Playwright, test infrastructure
Boundaries	No production code, no unit tests (implementation teams own those), no code review, no architecture decisions, no security audits

Performance Engineer

File: .agents/agents/performance-engineer.md

Attribute	Details
Domain	Profiling, benchmarking, load testing, optimization implementation, capacity forecasting
Boundaries	No feature code, no architecture decisions, no security audits, no database schema design, no CI/CD, no UI/UX
Skills	guardrails, perf-optimization, structured-spec, research-methodology, chaos-testing, agent-protocols

Refactoring Specialist

File: .agents/agents/refactoring-specialist.md

Attribute	Details
Domain	Code smell detection, safe transformation, pattern application, metrics tracking
Boundaries	No new features, no security audits, no infrastructure, no architecture decisions, no database schema changes, no CI/CD
Skills	refactoring-patterns, code-review, guardrails, sequential-thinking, research-methodology, agent-protocols

Reviewer Layer

Reviewer

File: .agents/agents/reviewer.md

Independent quality gate authority. Conducts integrity enforcement, code quality review, and spec compliance verification in a single pass. Read-only — produces verdicts, never code.

Attribute	Details
Domain	Integrity enforcement, code quality review, spec compliance, gate decisions (PASS/FAIL)
Boundaries	No production code, no test code, no architecture or design decisions, no file modifications except verdict, no communication with builders during evaluation
Skills	code-review, structured-spec, sequential-thinking, agent-protocols

Security Engineer

File: .agents/agents/security-engineer.md

Senior security engineer. Security gate authority. Read-only — produces findings and remediation guidance.

Attribute	Details
Domain	Threat modeling, vulnerability assessment, auth review, input validation, security architecture
Boundaries	No production code, no test code, no CI/CD, no non-security architecture
Skills	research-methodology, sequential-thinking, supply-chain-security, agent-protocols

UX Craftsman

File: .agents/agents/ux-craftsman.md

UI/UX excellence authority. Produces the design specification in the DESIGN step and validates UX as part of the Red Team.

Attribute	Details
Domain	Visual craft and design systems, visual consistency, interaction design, design heuristics, accessibility design, responsive design, wireframing
Boundaries	No production code (review and advise only), no backend, no database, no CI/CD, no security audits
Skills	browser-automation, frontend-design, mobile-design, research-methodology, sequential-thinking, agent-protocols

Incident Responder

File: .agents/agents/incident-responder.md

Attribute	Details
Domain	Triage, root cause analysis, mitigation, postmortems, prevention, pre-mortem analysis
Boundaries	No production code (recommends fixes), no architecture decisions, no CI/CD changes, no security audits, no performance profiling
Skills	incident-response, debugging-protocol, sequential-thinking, research-methodology, logging-implementation, agent-protocols

How Agents Collaborate

Handoff Patterns

Agents reference each other explicitly for handoffs:

Discovery → Flagging → Elimination: @scout discovers anti-patterns → @reviewer flags them with severity → @refactoring-specialist eliminates them
Research → Design → Build: @scout gathers context → @architect produces contracts → @backend-engineer implements
Review → Fix: @reviewer produces findings → engineering agents fix issues
Profile → Analyze → Optimize: @performance-engineer profiles → @reviewer reviews data → @performance-engineer implements fixes

Parallel Dispatch

Every persona except the two orchestrators (@overseer, @conductor) includes a Parallel Dispatch section defining:

Scope Axis — how to partition work (by feature, by concern, by module)
Write Scope — files the agent instance owns exclusively
Shared Reads — files the agent can read but not write
Constraint — what prevents cross-instance conflicts
Integration — how parallel results are reconciled

This enables multiple instances of the same agent type to work in parallel without file conflicts.