Use this agent when you need rapid development of complete applications, complex system architecture, performance-critical implementations, or when tackling challenging technical problems that require deep expertise across multiple technology stacks. This agent excels at building production-ready applications from scratch, optimizing existing codebases for maximum performance, and implementing sophisticated architectural patterns. Examples: - <example> Context: User needs a complete web application built quickly user: "I need a real-time chat application with user authentication" assistant: "I'll use the architect agent to rapidly build this complete chat application with all the features you need" <commentary> Since the user needs a full application built, use the architect agent to handle the complete implementation. </commentary> </example> - <example> Context: User has performance issues in their codebase user: "This API endpoint is taking 5 seconds to respond" assistant: "Let me bring in the architect agent to analyze and optimize this for maximum performance" <commentary> Performance optimization requires deep expertise, so the architect agent is perfect for this task. </commentary> </example>
Use this agent to merge parallel worktree branches into a single coherent codebase. The consolidator has full context of every subtask's intent and resolves conflicts by understanding code, not by picking sides. Launched automatically by the consolidation skill after parallel worktree agents complete. Can also be invoked directly when you have branches to merge and need intelligent conflict resolution. Examples: <example> Context: Multiple worktree agents have completed their subtasks and their branches need merging. user: "Merge these three feature branches together" assistant: "I'll use the consolidator agent to merge all branches with full context of what each one changed" <commentary> The consolidator agent has the merge protocol, conflict resolution framework, and post-merge verification to handle this cleanly. </commentary> </example> <example> Context: The consolidation skill has completed Phase 3 and needs to merge results. assistant: "Launching the consolidator agent to merge all worktree branches and wire everything together" <commentary> The consolidation skill triggers the consolidator agent for Phase 4 automatically. </commentary> </example>
Use this agent to decompose a task into the smallest reasonable units of work, decide which agents execute each unit, map file overlaps and dependencies, and produce an execution plan optimized for maximum parallelism. The planner is the first step in the consolidation cycle — it reads the codebase, understands the task, and outputs a structured plan that the verifier validates and the conducting agent executes. Examples: <example> Context: A multi-part feature needs to be implemented. assistant: "I'll use the planner agent to decompose this into parallel subtasks before launching any work" <commentary> The planner analyzes the codebase and produces an optimal execution plan with file ownership, agent assignments, and dependency graph. </commentary> </example>
Use this agent when you need an uncompromising code review that catches every possible issue, from critical bugs to minor style inconsistencies. This agent excels at identifying performance bottlenecks, security vulnerabilities, maintainability concerns, and code smells that others might overlook. Perfect for pre-production reviews, critical system components, or when you want to ensure your code meets the highest standards. Examples: <example> Context: The user wants a thorough code review after implementing a new feature. user: "I've just finished implementing the user authentication module" assistant: "I'll use the reviewer agent to perform a comprehensive review of your authentication implementation" <commentary> Since the user has completed a critical security feature, use the reviewer agent to ensure no vulnerabilities or issues exist. </commentary> </example> <example> Context: The user has written performance-critical code. user: "Here's my implementation of the sorting algorithm for our data pipeline" assistant: "Let me invoke the reviewer agent to scrutinize this performance-critical code" <commentary> Performance-critical code requires thorough review, so the reviewer agent should examine every aspect. </commentary> </example>
Use this agent to validate execution plans before work begins (pre-verification) and to confirm that completed work matches expected outcomes (post-verification). The verifier checks plans for correctness, completeness, efficiency, and effectiveness. It catches missing subtasks, unnecessary work, wrong file targets, and broken dependency chains before any agent writes code. Post-execution, it confirms the output matches the plan's acceptance criteria. Examples: <example> Context: The planner has produced an execution plan. assistant: "I'll use the verifier agent to validate this plan before launching any work" <commentary> The verifier catches issues in the plan before expensive parallel agents are spawned. </commentary> </example> <example> Context: All subtasks are complete and consolidated. assistant: "I'll use the verifier agent to confirm the output matches the expected outcome" <commentary> Post-execution verification ensures nothing was missed or implemented incorrectly. </commentary> </example>
House rules and load-bearing patterns for production Android with Jetpack Compose + MVI. Covers the MviViewModel<State,Intent,Effect> contract pattern, DispatcherProvider, collectAsStateWithLifecycle, type-safe Navigation Compose (@Serializable routes) + Navigation3 forward path, Koin DI (not Hilt), strong skipping defaults, edge-to-edge mandate, Material 3 / Expressive + adaptive layouts, Room + DataStore, Turbine + MockK + Robolectric testing scaffold, Baseline Profiles, R8 full mode, composable parameter ordering, module layout, and anti-pattern catalog. **Does not** re-teach elementary Compose (recomposition, remember, LaunchedEffect basics), standard Kotlin/coroutine patterns, or basic Material 3 usage -- assume that baseline. Use when writing or reviewing any Android/Compose code.
Backend API design, database architecture, microservices patterns, and test-driven development. Use for designing APIs, database schemas, or backend system architecture.
Build the project (auto-detects build system)
Find and remove dead code, unused imports, and technical debt
Create git commits in logical groups for all current changes
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Personal Claude Code commands, skills, and agents for global use across all projects.
Install via the built-in plugin marketplace:
/plugin marketplace add abnegate/claudes
/plugin install skills@claudes
Slash commands are namespaced under skills:, so they're invoked as /skills:commit, /skills:pr, etc.
Clone the repo and symlink into your Claude config — useful when iterating on the contents:
git clone [email protected]:abnegate/claudes.git ~/Local/claudes
ln -sf ~/Local/claudes/commands ~/.claude/commands
ln -sf ~/Local/claudes/skills ~/.claude/skills
ln -sf ~/Local/claudes/agents ~/.claude/agents
With symlinks, commands are invoked without the namespace prefix (/commit, /pr, ...).
Specialized agents that form a structured execution cycle. The cycle is conducted by the top-level agent (you, in Claude Code) via the consolidation skill — subagents cannot spawn further subagents, so the conductor role lives in the user-facing context.
[top-level agent — consolidation skill loaded]
-> planner (decompose task into subtasks)
-> verifier (validate plan correctness + efficiency)
-> architects (parallel worktree execution)
-> consolidator (merge all branches)
-> reviewer (review merged output)
-> verifier (confirm acceptance criteria met)
| Agent | Model | Role |
|---|---|---|
| planner | Opus | Decomposes tasks into smallest work units, maps dependencies and parallelism |
| verifier | Opus | Validates plans pre-execution and confirms outcomes post-execution |
| architect | Opus | Implements code in worktree isolation — production-ready, any tech stack |
| consolidator | Opus | Merges parallel worktree branches with intelligent conflict resolution |
| reviewer | Opus | Reviews code for bugs, security, performance, readability, and maintainability |
User-invocable slash commands. Type /<name> in Claude Code to run them.
| Command | Usage | Description |
|---|---|---|
| commit | /commit [message] | Create git commit with conventional commit message |
| commit-all | /commit-all | Create git commits in logical groups for all current changes |
| pr | /pr [title] | Commit pending changes, push, and create a pull request |
| pr-fix | /pr-fix <url> | Fix failing CI checks and address PR comments |
| issue | /issue <issue-id> | Implement a Linear issue end-to-end using the consolidation cycle |
| hotfix | /hotfix <description> | Emergency hotfix workflow for production issues |
| release | /release [version] [branch] | Create a GitHub release with grouped changelog |
| orchestrate | /orchestrate <description> | End-to-end feature workflow — branch, implement, improve, PR, wait, pr-fix |
| Command | Usage | Description |
|---|---|---|
| implement | /implement <feature> | TDD feature implementation via the consolidation cycle |
| refactor | /refactor <target> | Safe refactoring with planner, parallel worktrees, and verification |
| build | /build [target] | Build project (auto-detects build system) |
| install | /install [--device <target>] | Install app on device/emulator (auto-detects platform) |
| run | /run [--device <target>] | Build, install, and launch app on target device |
| Command | Usage | Description |
|---|---|---|
| improve | /improve [cycles] | Review and fix code across 6 dimensions — security, performance, correctness, readability, maintainability, testing |
| review | /review | Read-only code review of current branch against main |
| cleanup | /cleanup [module|all] | Remove dead code, unused imports, and technical debt |
| debug | /debug <error> | Debug and fix failing tests or errors |
| investigate | /investigate <issue> | Deep investigation of bugs, performance issues, or unexpected behavior |
| Command | Usage | Description |
|---|---|---|
| continue | /continue [context] | Pick up unfinished work from where the last session left off |
| history | /history <query> | Search Claude Code conversation history on disk |
| profile | /profile | Build a developer profile from git activity and session history |
| readme | /readme | Assess the codebase and update the README with any new or outdated information |
Reference guides loaded by Claude on demand. These are not user-invocable — Claude consults them automatically when relevant context appears.
npx claudepluginhub abnegate/claudes --plugin skillsComprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Write feature specs, plan roadmaps, and synthesize user research faster. Keep stakeholders updated and stay ahead of the competitive landscape.
Curate auto-memory, promote learnings to CLAUDE.md and rules, extract proven patterns into reusable skills.