Orchestrate Claude Code sessions as a Kanban board with agent lifecycle hooks, role-based task assignment, and automated verification of web pages, CLI tools, APIs, databases, and libraries, plus fuzzy exploration and plan decomposition into worker-ready tickets.
This skill should be used when the user asks to "QA browser changes", "verify a web UI", "test the browser flow", "check the page works", "validate the dev server", or wants end-to-end behavioral verification of web pages or web apps. Drives a real browser to navigate, snapshot, interact, and capture evidence — supports both agent-browser (preferred) and playwright-cli, picking whichever is installed.
This skill should be used when the user asks to "QA a CLI tool", "verify a command-line program", "check the binary's output", "validate exit codes", "compare against a golden file", or wants end-to-end behavioural verification of command-line tools. Drives a binary with representative inputs, captures stdout / stderr / exit code as separate evidence channels, asserts via golden-file comparison with volatile-field redaction.
This skill should be used when the user asks to "verify database state", "check what's in the database", "assert DB rows after the action", "validate the migration", or wants behavioural verification of database side effects. Snapshots database state before the action, exercises the system, snapshots after, and diffs structurally — ignoring volatile fields the spec does not assert. Used by primary-surface oracles (http-qa, cli-qa, library-qa) when a card asserts side-effect behaviour.
This skill should be used when the user asks to "QA an HTTP endpoint", "verify an API", "check the API behaves", "drive the API", "validate a route", or wants end-to-end behavioural verification of HTTP/JSON services. Drives a real running service with curl + jq, asserts on status / shape / headers / timing, captures the full request-response transcript as evidence, and distinguishes spec violations from flake.
This skill should be used when the user asks to "QA a library", "verify the package", "check the API behaves", "exercise the module", or wants end-to-end behavioural verification of an importable library or package (Go, Ruby, TypeScript, etc.). Writes a one-shot consumer program in scratch space, runs it against the library under test, observes structured output, and judges whether the behaviour matches the card's spec.
No model invocation
Executes directly as bash, bypassing the AI model
Runs pre-commands
Contains inline bash commands via ! syntax
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.
A ralph loop sets a target and works until it's done. ralph-ban makes that target a kanban board.
Terminal kanban built with Bubble Tea, backed by beads-lite's SQLite database. Five columns (Backlog, To Do, Doing, Review, Done) with vim-style navigation. The board pans horizontally when the terminal is too narrow. Cards sync across the bl CLI, Claude Code hooks, and other TUI sessions via 2-second polling.
My mate said I should have named this Gas-Ban, but it's not nearly that weird. Planner helps break down the tasks ralph-ban claude --plan. Orchestrator delegates to workers, reviewers, and oracles ralph-ban claude. Keep Orchestrator working until ToDo/Doing/Review are moved to Done ralph-ban claude --auto. That's about it.
Requires Go 1.25+.
go install github.com/kylesnowschwartz/ralph-ban@latest
Or build from source:
git clone [email protected]:kylesnowschwartz/ralph-ban.git
cd ralph-ban
go build .
Initialize a project, then open the board:
ralph-ban init --demo # create project with a demo board (Conway's Game of Life)
ralph-ban # open the board
ralph-ban ships a Claude Code plugin with hooks and agents. ralph-ban init extracts them into .ralph-ban/plugin/ and .claude/agents/.
ralph-ban claude # batch mode (pauses for human merge approval)
ralph-ban claude --auto # drain the board without intervention
ralph-ban claude --continue # continue most recent session
ralph-ban claude --resume # interactive session picker
? toggles full help. Ctrl+z suspends the TUI (resume with fg).
Navigation
| Key | Action |
|---|---|
h / ← | Focus left column |
j / ↓ | Move cursor down |
k / ↑ | Move cursor up |
l / → | Focus right column |
/ | Search cards |
Cards
| Key | Action |
|---|---|
n | New card |
e | Edit card |
d | Delete (press twice to confirm) |
z | Zoom card detail |
Enter | Move card right |
Backspace | Move card left |
u | Undo last move |
+ / - | Change priority |
Requires just for task running. Uses a go.work workspace with ../beads-lite for local development.
just build # build
just test # run tests
just run # build and launch TUI
just lint # vet + staticcheck
just release # tag, push, create GitHub release
MIT
npx claudepluginhub kylesnowschwartz/ralph-ban --plugin ralph-banUnified review, refactoring, and cleanup workflows with specialized analysis agents
Core SimpleClaude framework with intent-based commands and specialized agents
Skills and agents for diagrams, codebase patterns, hypothesis testing, Socratic thinking, file querying, frontend design, image generation, and command creation
Learn Claude Code by building real things, one level at a time
Enforces git diff thresholds to promote disciplined code review during AI agent sessions
Persistent Kanban board workflow with TodoWrite sync and SessionStart resume
Manage and decompose work into feature-level tasks on the agtx kanban board
Personal task tracker — invisible kanban that tracks your work automatically
Markdown-based Kanban board managed by Claude Code. Cards live as .md files — no database, no server.
Cotask — task management with TASKS.md kanban dashboard
Kaiten board automation tools