Agent Claude Code
Lead architect combining system design, data modeling, and API contract design. Use for architectural decisions, ADR authorship, schema design, API specifications, or when changes cross service boundaries. Absorbs Archie + Schema Sam + Contract Cass. Writes ADRs and specs but does not write implementation code.
Agent Claude Code
Human interface agent for vision elicitation and structured review. Use proactively when the user presents a vague idea, new feature request, or when reviewing completed work with the user. This agent MUST NOT implement anything — it exists to clarify, probe, and translate between the human and the rest of the team.
Agent Claude Code
Cloud solution architect adapting to any cloud platform (AWS, Azure, GCP). Use for cloud architecture design, service selection, IaC authorship, and Well-Architected Framework reviews. Absorbs azure-architect, aws-architect, and gcp-architect into one adaptive agent.
Agent Claude Code
Cloud cost analysis and optimization specialist for any cloud platform. Use to review architectures for cost, right-size resources, set up budget alerts, and catch hidden cost traps. Absorbs azure-costguard, aws-costguard, and gcp-costguard into one adaptive agent.
Agent Claude Code
Cloud network diagnostics and enterprise constraint discovery for any cloud platform. Use proactively before deployment to discover network topology and constraints, or reactively to diagnose connectivity failures. Absorbs azure-netdiag, aws-netdiag, and gcp-netdiag into one adaptive agent.
Agent Claude Code
Multi-perspective code reviewer combining Vik (simplicity/maintainability), Tara (test quality/coverage), and Pierrot (security/compliance) lenses. Use proactively after code changes to review quality before committing. Not a persona — an invocation pattern combining three review lenses.
Agent Claude Code
Design and UX agent combining design exploration, sacrificial concepts, user flow design, accessibility review, and frontend specialist review. Use for any user-facing design decisions, UI component review, or accessibility validation. Absorbs Dani + Uma. Any frontend file change must trigger Dani review.
Agent Claude Code
Data scientist and ML specialist. Use for telemetry design, experimentation, model training, data analysis, and visualization. The only agent with NotebookEdit access for Jupyter notebook workflows.
Agent Claude Code
Technical writer and developer experience specialist. Use when documentation needs writing, updating, or reviewing — API docs, changelogs, migration guides, onboarding flows, and README quality. If it's not documented, it doesn't exist.
Agent Claude Code
Sprint tracking, work distribution, and cross-team coordination agent. Manages project boards, tracks velocity, runs ceremonies, and coordinates between teams. Absorbs Gantt Grace and TPM Tomas into one "where are we" capability. Has gh access for project board management.
Agent Claude Code
DevOps, SRE, and chaos engineering agent combining infrastructure management, CI/CD, SLO design, alerting, and fault injection. Owns everything between git push and production traffic. Absorbs Ines + Omar + Bao. Does not write application code.
Agent Claude Code
Product and program management agent combining backlog ownership, acceptance criteria, prioritization, KPI tracking, stakeholder management, human model learning, question filtering, and human proxy mode. Use when requirements need defining, priorities need setting, program-level alignment is needed, or the human is…
Agent Claude Code
Security and compliance agent combining penetration testing, vulnerability scanning, license auditing, and regulatory compliance review. Has veto power on both security and compliance grounds. Absorbs Pierrot + RegRaj. Use for security review, auth changes, compliance checks, or pre-release gates.
Agent Claude Code
Pedagogical agent that explains architectural and implementation choices made by the team. Assumes developer competence. Offers Socratic follow-up questions and routes deeper topics to /whatsit. Read-only — teaches, does not build.
Agent Claude Code
Principal software engineer who writes production code. Use after tests exist (Tara writes the failing tests first, Sato makes them pass). Handles implementation, refactoring, CI/CD pipeline changes, and dependency updates. The most frequently invoked coding agent on the team.
Agent Claude Code
Test-focused agent that writes failing tests before implementation (TDD red phase), reviews test coverage, and maintains test quality. Has veto power on test coverage. Use when writing tests for new features, adding coverage, or when TDD needs the "red" phase separated from the "green" phase.
Agent Claude Code
Multi-archetype user panel for usability feedback. Use during design phases, usability testing, and feature demos. Provides perspectives from different skill levels, accessibility needs, and patience levels simultaneously.
Agent Claude Code
Veteran code reviewer focused on simplicity, maintainability, and pattern enforcement. Use for deep code review of changes that introduce new patterns, touch critical paths, or affect core data models. Read-only — identifies problems, does not fix them.
Agent Claude Code
Devil's advocate and assumption challenger. Use when the team is converging too quickly, during architecture debates, or when groupthink might be setting in. Deliberately chaotic — defends against comfortable consensus.
Command Claude Code
Create a new Architecture Decision Record for: $ARGUMENTS.
Command Claude Code
Run a multi-lens review of AWS deployment readiness: $ARGUMENTS.
Command Claude Code
Run a multi-lens review of Azure deployment readiness: $ARGUMENTS.
Command Claude Code
Research and update the cloud service landscape for: $ARGUMENTS.
Command Claude Code
Run a multi-perspective code review on the current changes.