By kobbikobb
A strict, opinionated wizard-driven workflow that wraps ralph-wiggum to plan and execute multi-milestone software projects using specialized agents.
Group a branch's diff into concept-level topics for refinement.
Answer the user's question about a refinement topic. Read-only.
Apply a refinement to a topic's files. Scope-bounded.
Implement a single milestone: write code, write tests, verify, and commit.
Run after the Planner, before save. Validate that each milestone is sized correctly and that the milestone graph is internally consistent.
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.
No framework bloat — just a wizard, a handful of agents, and a loop.
Most AI coding tools are either too simple (one-shot prompts) or too complex (30 commands, 20 agents, 7 modes). Vader sits in between:
/plugin marketplace add kobbikobb/vader
/plugin install vader@vader
Then:
/vader Add user authentication with JWT tokens
The wizard guides you through 5 stages, then:
/vader:exec
Walk away. Come back to committed, verified code.
Got an existing feature branch you want to review concept by concept?
/vader:refine
A Chunker groups the diff into topics. For each one: approve, discuss, edit, defer, jump, back, or skip. Edits go through an Editor and a Refine Verifier before a per-topic commit. Pushes to the PR at the end if one exists — never force.
/vader "description" /vader:exec
│ │
┌────▼────┐ ┌─────▼─────┐
│Research │ │ For each │
│codebase │ │ milestone:│
└────┬────┘ │ │
┌────▼────┐ │ Executor │──▶ implement + test
│ Draft │ │ │ │
│ plan │ │ Verifier │──▶ validate goal met
└────┬────┘ │ │ │
┌────▼────┐ │ Pass? ───│──▶ commit + next
│ Split │ │ Fail? ───│──▶ fix (up to 3x)
│mileston.│ └───────────┘
└────┬────┘
┌────▼────┐
│ Save │
│ plan │
└─────────┘
| Command | What it does |
|---|---|
/vader [description] | Plan a project — interactive wizard |
/vader:exec | Execute the plan autonomously |
/vader:refine | Walk the current branch's diff topic by topic |
/vader:status | Check progress |
/vader:cancel | Abort and clean up |
/vader:help | Usage guide |
Specialized agents, each a markdown file you can customize:
| Agent | Phase | Job |
|---|---|---|
| Researcher | Planning | Explores codebase, finds patterns, surfaces risks |
| Planner | Planning | Breaks project into dependency-ordered milestones |
| Executor | Execution | Implements code and tests for one milestone |
| Verifier | Execution | Validates the milestone goal was actually achieved |
| Chunker | Refinement | Groups a diff into concept-level topics |
| Discusser | Refinement | Answers questions about a topic (read-only) |
| Editor | Refinement | Applies refinements within a topic's file scope |
| Refine Verifier | Refinement | Checks an edit stayed in scope, no regressions |
Edit agents/*.md to match your team's conventions.
--dangerously-skip-permissions for uninterrupted overnight execution/vader:status then /vader:exec to resumeagents/ to encode your project's rulesMIT
npx claudepluginhub kobbikobb/vader --plugin vaderVibe Better With Claude Code - You're not an engineer anymore. You're a prompt jockey with commit access. At least do it properly.
Token-efficient vibe-coding pipeline with hard guardrails: brainstorm, plan, isolate, exec, verify, review, integrate. Halts on 'Expected' mismatches instead of committing bad work.
Compound Engineering workflow: PRD-driven sprints, isolated worktrees, hook-enforced safety, automated learning. Skills become /vini-workflow:plan, /vini-workflow:compound, etc.
Autonomous plan execution with Claude Code - task execution, monitoring, and plan creation
Zero-dependency planning + execution with .flow/ task tracking and Ralph autonomous mode (multi-model review gates). Worker subagent per task for context isolation. Prime assesses 8 pillars (48 criteria) with GitHub API integration. Includes 20 subagents, 11 commands, 16 skills.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses