Agents Reference

Agents are specialized subagents that clarc delegates work to. Most activate automatically — you don't need to invoke them manually.

Core Workflow Agents

Agent	Activates when	Manual command
`planner`	Complex features, refactoring	`/plan`
`tdd-guide`	New features, bug fixes	`/tdd`
`code-reviewer`	After writing or modifying code	`/code-review`
`security-reviewer`	Code touches auth, input, APIs, secrets	`/security-review`
`architect`	Architectural decisions	`/arc42`
`build-error-resolver`	Build fails	`/build-fix`
`e2e-runner`	Critical user flows need testing	`/e2e`
`refactor-cleaner`	Dead code cleanup	`/refactor`
`doc-updater`	Documentation needs updating	`/update-docs`

These are invoked automatically by code-reviewer based on file extension, or directly:

Agent	Files	Specialization	Manual command
`typescript-reviewer`	`.ts`, `.tsx`, `.js`	Type safety, DDD, hexagonal arch	`/typescript-review`
`go-reviewer`	`.go`	Idiomatic Go, error handling, concurrency	`/go-review`
`python-reviewer`	`.py`	PEP 8, type hints, Pythonic idioms	`/python-review`
`java-reviewer`	`.java`	Spring Boot, JPA, hexagonal arch	`/java-review`
`swift-reviewer`	`.swift`	SwiftUI, concurrency, protocol arch	`/swift-review`
`rust-reviewer`	`.rs`	Ownership, async/Tokio, error handling	`/rust-review`
`ruby-reviewer`	`.rb`	Rails patterns, Brakeman, N+1 detection	`/ruby-review`
`elixir-reviewer`	`.ex`, `.exs`	OTP patterns, Ecto, Sobelow	`/elixir-review`
`kotlin-reviewer`	`.kt`	Coroutines, null safety, Compose	`/kotlin-review`
`cpp-reviewer`	`.cpp`, `.h`	C++20, RAII, memory safety	`/cpp-review`
`csharp-reviewer`	`.cs`	C# 12/.NET 8, nullable types, async/await	`/csharp-review`
`php-reviewer`	`.php`	PHP 8.4, strict types, Laravel/Symfony	`/php-review`
`scala-reviewer`	`.scala`	Cats Effect/ZIO, ADTs, functional idioms	`/scala-review`
`r-reviewer`	`.r`, `.R`	tidyverse, purrr, Shiny, renv	`/r-review`
`bash-reviewer`	`.sh`, `.bash`	shellcheck, set -euo pipefail, BATS	`/bash-review`
`c-reviewer`	`.c`, `.h`	C11/C17, memory safety, opaque pointers	`/c-review`
`android-reviewer`	`.kt` (Android)	Jetpack Compose, Hilt, Room, ViewModel	`/android-review`
`flutter-reviewer`	`.dart`	Flutter architecture, const, RepaintBoundary	`/flutter-review`
`database-reviewer`	`.sql`, `.prisma`	PostgreSQL, query optimization, RLS	`/database-review`

Agent	Purpose	Invoked by
`product-evaluator`	Go/No-Go for product ideas	`/evaluate`
`solution-designer`	2–4 solution options + ADR	`/explore`
`competitive-analyst`	Competitor feature matrix	`/competitive-review`
`feedback-analyst`	Cluster themes from user feedback	`/analyze-feedback`

Agent	Purpose	Invoked by
`design-critic`	Structured visual critique — composition, hierarchy, color, typography, accessibility	`/design-critique`
`design-system-reviewer`	Comprehensive design system audit — tokens, dark mode, icons, accessibility, component completeness	manual
`presentation-designer`	Slide deck creation and structure — speaker notes, narrative, slide density	`/slide-deck`
`talk-coach`	Talk and presentation review — structure, timing, audience fit	`/talk-outline`

Agent	Purpose
`gitops-architect`	ArgoCD/Flux setup design
`mlops-architect`	ML serving, drift, A/B testing
`finops-advisor`	Cloud cost optimization
`frontend-architect`	Micro-frontend architecture
`data-architect`	Data Mesh design
`sdk-architect`	SDK architecture and release process
`devsecops-reviewer`	SAST, IaC misconfig, OWASP
`supply-chain-auditor`	SBOM, unpinned deps, unsigned artifacts
`performance-analyst`	Flamegraphs, N+1, profiling output
`resilience-reviewer`	Circuit breakers, retry patterns

These analyze and improve clarc itself:

Agent	Purpose	Invoked by
`agent-quality-reviewer`	Review a single agent file — scores across 8 dimensions	`/agent-audit`
`agent-system-reviewer`	Full system review — orchestrates all analyzers	`/system-review`
`command-auditor`	UX ergonomics audit across all commands	`/command-audit`
`hook-auditor`	Hook coverage, conflicts, and dead references	`/hook-audit`
`prompt-quality-scorer`	Rate all agent/command prompts for quality	manual
`skill-depth-analyzer`	Analyze a skill for prompt-engineering quality	`/skill-depth`

In Claude Code, ask Claude to use a specific agent:

Use the security-reviewer agent to check my authentication module.

Use the planner agent to create an implementation plan for adding OAuth.

Or use the relevant slash command, which invokes the agent automatically.