not-my-reforge

Autonomous execution mode — break a goal into tasks and execute them sequentially with safety checkpoints.

Deep analysis with opus model. Use for complex debugging, architectural decisions, difficult bugs, security analysis, and tasks requiring careful reasoning.

Deep project initialization — generate hierarchical AGENTS.md, CLAUDE.md, and directory-level documentation.

Multi-perspective parallel code review — analyze git diff from 5 viewpoints simultaneously.

Diagnose plugin health — check dependencies, hooks, agents, and runtime state.

Generate a session handoff document capturing current work state, progress, and context for the next session or developer.

View session history — past sessions, duration, tool usage, and trends.

Auto-generate CLAUDE.md by analyzing project structure, dependencies, conventions, and patterns. Creates a comprehensive project instruction file.

Ralph Loop — persistent iterative execution with automatic verification, failure adaptation, and completion detection. Keeps running until the task passes all checks or max iterations are reached.

Mode management — show active mode, mode history, and manage mode state.

Configure notification channels for session completion alerts (Discord, Slack, Telegram).

Multi-stage quality pipeline — Plan, Implement, Verify, Fix, Review with automated gate checks.

Generate PR metadata — title, description, changed files summary, and test plan.

Auto QA loop — run tests, analyze failures, fix, and retest automatically.

Fast execution with haiku model. Use for simple questions, quick lookups, small edits, and tasks where speed matters more than depth.

Iterative consensus planning — structured plan-critique-revise cycles until agreement.

Release management — version bumping, changelog generation, and release notes.

Code review based on git diff. Reviews staged/unstaged changes or a specific commit range for bugs, security issues, and code quality.

Decompose a complex task into parallel subtasks and generate tmux commands for concurrent execution with separate worktrees.

Display session metrics dashboard — tool calls, context usage, active modes, and notification status.

Swarm mode — dispatch multiple parallel agents for research, analysis, or bulk operations.

Create an Agent Teams setup with N workers + reviewer for parallel task execution with worktree isolation and merge coordination.

Ultrawork — activate ALL available specialized agents in maximum parallel execution for comprehensive task coverage.

Frontend and UI specialist. Use for React/Vue/Svelte components, CSS/Tailwind styling, accessibility, responsive design, and visual implementation tasks.

Requirements analyst — decomposes goals into specifications, identifies constraints, maps dependencies, and defines acceptance criteria.

Build and type error specialist — diagnoses and fixes compilation errors, type mismatches, and dependency issues.

Iterative consensus planning — planner and critic debate until agreement.

Plan critic — evaluates implementation plans for gaps, risks, over-engineering, and overlooked alternatives.

Multi-strategy codebase search agent — combines text, file, git, and symbol search for comprehensive code discovery.

UI/UX design specialist — component architecture, design system patterns, accessibility, and responsive layout.

Fast codebase explorer using haiku model. Use for quick file discovery, pattern searching, and rapid codebase orientation when speed matters more than depth.

Git operations specialist — branch strategy, conflict resolution, history analysis, and workflow automation.

Documentation and OSS research specialist. Use for finding API docs, library examples, best practices, migration guides, and reference implementations from the web.

Deep architecture analyst using opus. Use for thorough codebase analysis requiring complex reasoning — cross-module impact, architectural debt, migration planning.

Fast architecture scan using haiku. Use for quick structural questions, simple dependency checks, and rapid file lookups where deep analysis is unnecessary.

Read-only architecture analyst. Use for deep codebase analysis, dependency mapping, impact assessment, and architectural review without modifying any files.

Quick implementation planner using sonnet. Use for straightforward tasks that need a brief plan — simple features, bug fixes, small refactors.

Implementation planner using opus model. Use for complex task decomposition, architectural planning, multi-step implementation strategies, and risk analysis before coding.

Automated QA specialist — test execution, failure analysis, and fix iteration.

Fast code review using haiku. Use for quick sanity checks, simple diffs, and rapid feedback on small changes.

Code review specialist. Use for reviewing git diffs, pull requests, code quality assessment, security audit, and identifying bugs before merge.

Security vulnerability scanner — OWASP Top 10, dependency audit, secrets detection, and threat modeling.

Team lead agent for Agent Teams. Coordinates task decomposition, work assignment, and result integration across team members.

Read-only research agent for Agent Teams. Runs in background to gather information without modifying the codebase.

Code reviewer for Agent Teams. Reviews teammate changes without modifying code.

Test writer and runner for Agent Teams. Creates and executes tests in isolated worktrees.

Implementation worker for Agent Teams. Executes assigned tasks in isolated worktrees with full read-write capabilities.

Testing specialist — unit tests, integration tests, TDD workflow, mocking strategies, and test coverage analysis.

Visual analysis specialist — screenshots, mockups, UI comparisons, error screenshots, and design implementation guidance.

Multi-perspective code audit. Launches parallel security, quality, performance, and architecture analysis agents.

Multi-strategy parallel codebase search. Combines text search, file patterns, git history, and import tracing to find code comprehensively.

Automatically detect and understand project structure, tech stack, conventions, and key files. Provides project context as background knowledge.

Experiment with code changes in an isolated worktree. Changes can be reviewed and merged or discarded.

5-stage quality pipeline for Agent Teams — Plan, PRD, Execute, Verify, Fix

49 hooks across 9 events