By izmailovilya
Interactively audit React/TypeScript projects for dead code, unused features, tRPC routes, Zustand stores, and UI components through developer conversations. Receive structured JSON reports on suspicious items, review findings, confirm actions, and safely execute cleanups with git backup branches, TypeScript checks, and detailed commit reports.
npx claudepluginhub izmailovilya/ilia-izmailov-plugins --plugin vibe-auditExecutes cleanup actions after user confirms what to delete. Handles safe removal with git backup. <example> Context: User confirmed features to delete user: "Удали rat-hypothesis" assistant: "Запускаю cleanup-executor для безопасного удаления с git backup" </example>
Scans codebase for potentially unused or experimental features. Returns structured list for interactive review. <example> Context: User wants to clean up their vibe-coded project user: "Найди мёртвый код" assistant: "Запускаю feature-scanner для поиска потенциально неиспользуемых фич" </example>
This agent should be used when auditing src/features/ directory for dead code, unused exports, and duplication. Triggers on: "audit features", "find dead code in features", "clean up features/", "check features for unused code". <example> Context: User wants to audit feature modules for cleanup user: "Проверь features на мёртвый код" assistant: "Запускаю features-auditor для анализа неиспользуемых экспортов и дубликатов" </example>
This agent should be used when auditing src/server/ directory for dead code, API inconsistencies, and unused routers. Triggers on: "audit server", "find dead code in server", "clean up routers", "check API for unused endpoints". <example> Context: User wants to clean up server-side code user: "Проверь сервер на мёртвый код" assistant: "Запускаю server-auditor для аудита tRPC роутеров и сервисов" </example>
This agent should be used when auditing src/stores/ directory for unused stores, redundant state, and Zustand patterns. Triggers on: "audit stores", "find unused stores", "clean up state management", "check Zustand stores".
This agent should be used when auditing src/design-system/ and UI components for unused components, style inconsistencies, and duplication. Triggers on: "audit design system", "find unused components", "clean up UI", "check design-system for dead code".
Deep analysis of a specific feature's usage across the codebase. Called when user needs more details before deciding. <example> Context: User said "не уверен" about a feature user: "Расскажи подробнее про rat-hypothesis" assistant: "Запускаю usage-analyzer для детального анализа использования" </example>
A collection of plugins for Claude Code.
Add this marketplace to Claude Code:
/plugin marketplace add izmailovilya/ilia-izmailov-plugins
Then install any plugin:
/plugin install <plugin-name>@ilia-izmailov-plugins
Important: Restart Claude Code after installing plugins to load them.
Launch a team of AI agents to implement features with built-in code review gates.
Requires: Enable
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMSin settings.json or environment. See setup →
/plugin install agent-teams@ilia-izmailov-plugins
Usage:
/interviewed-team-feature "Add user settings page"
/team-feature docs/plan.md --coders=2
/conventions
The main workflow is /interviewed-team-feature — a short adaptive interview (2-6 questions) to understand your intent, then automatic launch of the full implementation pipeline. Spawns researchers, coders, and specialized reviewers (security, logic, quality) with automatic team scaling based on complexity (SIMPLE/MEDIUM/COMPLEX).
Expert evaluation arena — real experts independently assess options with cross-enrichment for any domain.
Requires: Enable
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMSin settings.json or environment. See setup →
/plugin install expert-arena@ilia-izmailov-plugins
Usage:
/expert-arena "Should we use microservices or monolith?"
/expert-arena "Best pricing strategy for a developer tool?"
Selects 3-5 real experts with opposing viewpoints, gathers context via researchers, launches independent evaluations with cross-enrichment, and produces an action-oriented report: verdict first, action plan second, detailed analysis for those who want to dig deeper.
Deep parallel codebase research — causal understanding, not just coverage.
Requires: Enable
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMSin settings.json or environment.
/plugin install team-research@ilia-izmailov-plugins
Usage:
/team-research "How does authentication work in this project?"
/team-research "Full architecture review"
Spawns a scout to map the landscape, then 2-7 investigators explore independent angles in parallel, followed by an adversarial challenger who stress-tests the findings. Produces a research report with causal understanding, source confidence tags, and cross-cutting insights.
Scout open-source repos for patterns and ideas to improve your own product.
/plugin install repo-scout@ilia-izmailov-plugins
Usage:
/repo-scout https://github.com/anomalyco/opencode
/repo-scout https://github.com/vercel/ai "how they handle streaming"
Two-phase approach: first understands YOUR project (2 scouts), then explores the external repo with your context (2 scouts), then adversarial challenge (2 challengers verify patterns are real and worth adopting). Only recommendations that survive the challenge make it into the final report.
Interactive feature audit for vibe-coded projects. Finds dead code, unused features, and experiments through conversation.
/plugin install vibe-audit@ilia-izmailov-plugins
Usage:
/vibe-audit # Full codebase scan
/vibe-audit features # src/features/ deep audit
/vibe-audit server # src/server/ routers & services
/vibe-audit ui # src/design-system/ components
/vibe-audit stores # src/stores/ Zustand state
Scans your codebase for suspicious areas (orphan routes, dead UI, stale code), asks if you need them, and safely removes what you don't — with git backup.
MIT
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Adversarial multi-agent pipeline for Claude Code. GAN-style loops where generators produce artifacts, discriminators validate them, and feedback drives convergence.
24 parallel audit agents + 6 workflow skills for Claude Code. Complete E2E development workflow for solo devs.
A caring skill pack for vibe coders. Encodes engineering discipline — refactor verification, minimal security audit, repo rot detection, AI-friendly docs, CI/CD basics, secrets lifecycle, project conventions, repo bloat diagnosis, and design-system unification — as a set of portable agent skills.
コードベースの不要なフォールバック・互換性コード・デッドコードを 3つの専門エージェント（scanner + historian + verifier）が協調して検出・安全性検証する監査スキル。 scanner が自律探索で候補を発見し、historian が git 履歴から時間軸の文脈を付与し、
Code review plugin with a standalone reviewer agent and two skill strategies: disposable subagents for one-shot reviews and persistent team members for iterative reviews