Open-source Claude Code plugins by Roxabi — context engineering tools for teams using Claude Code.
Agent
Red-team / devil's advocate for specs, diffs, analyses, proposals, architecture, and ideas. Attacks assumptions, control effectiveness, vacuous guards, fleet impact, and partial-failure paths. On /dev-review: also apply an OWASP lens (secrets, injection, auth) — security-auditor is not spawned by default. Invoked by…
Agent
Use this agent for system design decisions, cross-cutting architecture, and technical planning across the monorepo. Context: New feature requires architectural decisions user: "Design the caching strategy for the API" assistant: "I'll use the architect agent to design the architecture.".
Agent
Interview agent for the foundational "Axis of Decomposition" ADR — captures the primary axis of system variation to prevent N×M drift (cross-cutting concerns duplicated across non-primary axis siblings). Invoked by /init Phase 3a (mandatory) or standalone (re-run, supersede). Conducts 4 mandatory + 1 optional question…
Agent
Read-only review agent for axial-decomposition drift. Parses the project's axial ADR (axial: true frontmatter) and reviews a diff or spec for drift along the non-primary axis (N×M trap). Emits Conventional Comments findings tagged with the canonical target-axis-trap class. Invoked by /spec (Step 4 — Expert Review) or…
Agent
Use this agent for backend implementation tasks in any framework and ORM. Works with NestJS, Express, Fastify, Django, Rails, and any ORM. Context: User needs a new API endpoint user: "Create an endpoint to fetch user preferences" assistant: "I'll use the backend-dev agent to implement the API.".
Agent
Use this agent for infrastructure, CI/CD, dependency management, and configuration tasks. Works with any build orchestrator, formatter, package manager, and deployment platform. Context: CI/CD pipeline issue user: "GitHub Actions is failing on the typecheck step" assistant: "I'll use the devops agent to debug the CI…
Agent
Use this agent for documentation creation, maintenance, and CLAUDE.md updates. Default write format is plain Markdown (.md). Legacy .mdx is readable but not preferred for new files. Context: New feature needs documentation user: "Document the new auth module" assistant: "I'll use the doc-writer agent to create the…
Agent
Use this agent to fix specific review comments across the entire stack (frontend, backend, tests, config). Receives accepted review findings and applies targeted fixes without writing new features or refactoring beyond what is needed. Context: Review comments have been accepted by the human user: "Fix these accepted…
Agent
Use this agent for frontend implementation tasks in any framework and UI library. Works with TanStack Start, Next.js, Remix, SvelteKit, Nuxt, and any UI component system. Context: User needs a new page or component implemented user: "Implement the user profile page" assistant: "I'll use the frontend-dev agent to…
Agent
Use this agent for product leadership: requirements gathering, issue triage, prioritization, writing analyses and specs, driving the dev pipeline, and verifying deployed features. Context: New feature needs requirements user: "Gather requirements for the notification system" assistant: "I'll use the product-lead agent…
Agent
Targeted recall agent for cross-chunk class join in /dev-review. Receives only callsites + context window, never the full diff. Confirms scope (RC-3) and finds un-cited instances (RC-6) for a single class. Context: class "parallel-path-drift" hit in 3 chunks orchestrator: "Spawn recall agent for parallel-path-drift"…
Agent
Audit code for security vulnerabilities, review auth flows, and verify compliance with OWASP top 10. Context: Security review needed before release user: "Audit the authentication module for vulnerabilities" assistant: "I'll use the security-auditor agent to perform a security audit.".
Agent
Use this agent to generate tests, validate coverage, and verify test quality. Works with Vitest, Jest, Pytest, Playwright, Cypress, and any test framework. Context: New feature needs tests user: "Write tests for the auth service" assistant: "I'll use the tester agent to generate test coverage.".