By fintanislost
Auto-triggering skill + slash commands for kadai (the local-first product spine for projects driven by agentic coding)
Create a kadai item (epic, feature, story, or task) with prompts for missing fields
Disable kadai in the current project. Hooks no-op, mutating CLI commands error, MCP mutating tools refuse, web viewer shows a DISABLED banner. Reads still work.
Re-enable kadai in the current project. Hooks resume gating, mutating commands work again, web banner clears.
Pick a kadai story for active work (sets it as picked AND transitions to in_progress)
Render all descendant story plans of an epic, feature, or story as one composite markdown document.
Use INSTEAD OF superpowers:brainstorming when working in any repository that contains a `.kadai/` directory. Runs the full brainstorming flow (explore → questions → approaches → design → user approval → spec → self-review) AND then plumbs the result into the kadai spine: auto-creates the epic if missing, creates a feature, attaches the spec to the feature as `spec.md`. The single visible artifact at the end is a feature in `.kadai/epics/<E>/features/<F>/` with `spec.md` attached, NOT a free-floating `docs/superpowers/specs/*.md`. If you load `superpowers:brainstorming` in a kadai repo, you'll write specs into the wrong place and lose the spec → plan → implementation provenance kadai is designed to capture.
Use when verifying that kadai's full integration (skill auto-trigger + PreToolUse/PostToolUse hooks + MCP server + change capture + status updates) works end-to-end. Spawns a fresh `claude -p` session in a temp dir to exercise the Path A acceptance test from the kadai design spec §10. Useful after kadai upgrades, install changes, or when something feels off about hooks/MCP/skill behavior.
Autonomous executor for kadai-tracked work. Picks up where kadai-writing-plans leaves off — walks the picked story's plan task-by-task, dispatches an implementer subagent for each ### Task block, runs spec + code-quality reviews per task, marks tasks done as it goes. Pauses for human review at story boundaries. When an implementer reports BLOCKED with reason `needs-feature: <desc>`, surfaces a prompt to plan a fast-follow-up unblocker via kadai-brainstorming, executes the unblocker, then resumes the original story. State persisted in `.kadai/runner.json`; resumable across sessions via `kadai run --resume`.
Use INSTEAD OF superpowers:writing-plans when writing implementation plans inside a kadai-tracked repo (`.kadai/` present). Decomposes the feature into kadai stories, writes one `plan.md` per story directly into the spine (`.kadai/.../stories/<id>/plan.md`), and creates one real kadai task for each `### Task N` block. Auto-picks the first story when done. Without this wrapper you'll produce a single freeform plan that someone has to translate into spine items by hand.
Required when working in any repository that contains a `.kadai/` directory — including brainstorming, designing, scoping, planning, building, or implementing anything (MVPs, features, prototypes, fixes, websites, scripts). Load this skill at the START of the session, before invoking brainstorming/writing-plans or writing any files. Kadai enforces a guardrail (edits outside the spine + allowlist are blocked when no story is picked) and tracks the spec→plan→implementation trail in `.kadai/`. If you skip this skill, you'll write specs/plans into the wrong place and lose the provenance kadai is designed to capture.
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.
Local-first product spine for projects driven by agentic coding.
Kadai tracks a hierarchical scope — epics → features → stories → tasks — as markdown files committed to your repo, and exposes typed MCP tools + Claude Code hooks that keep agents honest to the spine. Agents can't edit code that doesn't map to a picked story; every edit they do make is captured to that story's changelog.
Three things ship in this repo:
kadai) — manage the spine from the terminal.kadai mcp) — exposes the spine to agents through typed tools.kadai serve) — roadmap + drill-down views at localhost, with live updates and spine-wide search.Plus a bundled Claude Code plugin (kadai-plugin/) — auto-triggering skill and slash commands (/kadai-pick, /kadai-status, /kadai-disable, …).
From source (requires Bun ≥ 1.1):
git clone https://github.com/fintanislost/kadai.git
cd kadai
bun install
bun run build:web
bun link # exposes the `kadai` command
Or build a self-contained binary:
bun run build # writes dist/kadai for your host
bun run build:all # cross-compiles for 5 targets
Other install paths (curl, Homebrew, npm) are documented in docs/wiki/installation.md. Releases are pending — until then, install from source.
cd ~/projects/my-app
kadai init # creates .kadai/, .mcp.json, hooks
kadai add feature --title 'User login' --phase mvp --parent EPIC-001
kadai add story --title 'Email login' --phase mvp --parent FEAT-001
kadai pick STORY-001 # picks + transitions to in_progress
kadai status # picked story + queue
kadai serve # roadmap at localhost
Restart your Claude Code session in the project directory after kadai init so the MCP server and hooks load.
To install the Claude Code plugin (so agents auto-trigger the kadai skill):
/plugin marketplace add /path/to/kadai-repo
/plugin install kadai@kadai
/reload-plugins
After kadai pick <story>:
Edit / Write outside .kadai/ and the configured allowlist when no story is picked.changelog.md.kadai disable turns the guardrail off project-wide (or globally / per env var); kadai enable brings it back. See docs/wiki/concepts.md for the toggle semantics.User-facing documentation lives in docs/wiki/:
bun test # full test suite
bun test src/core # core module only
bun test src/cli # CLI only
Tech stack: TypeScript on Bun (runtime + test runner + bundler), commander (CLI), zod (validation), gray-matter (frontmatter), smol-toml (config), @modelcontextprotocol/sdk (MCP), Vite + React + Tailwind + TanStack Router (web viewer).
TBD.
npx claudepluginhub fintanislost/kadai --plugin kadaiComprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Complete developer toolkit for Claude Code
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Orchestrate multi-agent teams for parallel code review, hypothesis-driven debugging, and coordinated feature development using Claude Code's Agent Teams