Agent skills for AI coding assistants. Install via skills.sh.
Install all skills:
npx skills add laststance/skillsInstall a specific skill:
npx skills add laststance/skills --skill apollousa
npx skills add laststance/skills --skill chrome-clean-install
npx skills add laststance/skills --skill ci-hardening
npx skills add laststance/skills --skill claude-code-plugin-troubleshoot
npx skills add laststance/skills --skill code-trace
npx skills add laststance/skills --skill codebase-litter-audit
npx skills add laststance/skills --skill coderabbit-resolver
npx skills add laststance/skills --skill colorful-type
npx skills add laststance/skills --skill component-hierarchy
npx skills add laststance/skills --skill cookie
npx skills add laststance/skills --skill create-worktree
npx skills add laststance/skills --skill deep-trace
npx skills add laststance/skills --skill dnd
npx skills add laststance/skills --skill electron-release
npx skills add laststance/skills --skill explain
npx skills add laststance/skills --skill explain-diff-html
npx skills add laststance/skills --skill github-actions-pnpm-ci
npx skills add laststance/skills --skill laststance-publish-skill
npx skills add laststance/skills --skill learn-html
npx skills add laststance/skills --skill load
npx skills add laststance/skills --skill locate-ui-from-code
npx skills add laststance/skills --skill pr-tour
npx skills add laststance/skills --skill product-inspiration
npx skills add laststance/skills --skill prop-drill
npx skills add laststance/skills --skill react-query-key-jump
npx skills add laststance/skills --skill rec
npx skills add laststance/skills --skill save
npx skills add laststance/skills --skill search
npx skills add laststance/skills --skill source-grounded-research
npx skills add laststance/skills --skill ts-pattern-refactor
npx skills add laststance/skills --skill type-expand
npx skills add laststance/skills --skill ux-gap-detector
npx skills add laststance/skills --skill video
npx skills add laststance/skills --skill visual-lint| Skill | Description | Dependencies |
|---|---|---|
| apollousa | Creates a GitHub PR for completed work, then runs the full CodeRabbit review, CI, merge, and cleanup loop. | coderabbit-resolver (required) |
| chrome-clean-install | Refresh Chromium-based browsers by backing up profile/cache data, guiding a clean reinstall, and restoring bookmarks only. Handles Chrome, Chrome Canary, Edge, Brave, Arc, Dia, and custom Chromium browser paths. | Node.js (required) |
| ci-hardening | Port the skills-desktop GitHub Actions hardening baseline — CodeQL, Dependency Review, Scorecard, Dependabot, CODEOWNERS, pinned actions, least-privilege permissions, and branch/security settings. | Node.js (required), GitHub CLI (recommended) |
| claude-code-plugin-troubleshoot | Debug, audit, and fix Claude Code plugin system issues — hook errors, plugin misbehavior, cache investigation. Knows that enabledPlugins: false is not a true kill switch (hooks still execute, skills still accessible). |
— |
| code-trace | Interactive code execution path tracer. Explains how code flows from entry point to output with step-by-step navigation. | — |
| codebase-litter-audit | Audit repositories for half-finished features, stale TODOs, no-op handlers, visible UI wired to stubs, disabled tests, stale docs, placeholder assets, suppressions, and other codebase litter that dead-code tools miss. | — |
| coderabbit-resolver | Audits inline and outside-diff CodeRabbit findings, validates fixes, rechecks review evidence before merge, and cleans up. Supports --bulk for all open PRs. |
GitHub CLI (required), jq (required) |
| colorful-type | Replace colorless primitives (string, number, boolean) with domain-rich types. Adds branded types, JSDoc, and named type aliases to communicate intent. |
— |
| component-hierarchy | Visualize where a React component sits in the Next.js tree (Page → target) as an ASCII diagram with file paths. Supports App Router and Pages Router. | — |
| cookie | Copy Google Chrome's cookies into playwright-cli (macOS) so its browser inherits every logged-in session (GitHub, etc.). Decrypts via the macOS Keychain, loads per-cookie, verifies before navigating, and deletes the plaintext token files after. |
playwright-cli (required), Node.js (required) |
| create-worktree | Creates a git worktree as a sibling directory to the current project (e.g., ../project-feat-x), copies .gitignored config files (.env, .env.local, etc.) while skipping heavy build/dependency directories (node_modules, .next, dist, build, coverage), then navigates into the new worktree. |
— |
| deep-trace | Line-by-line execution path tracer for PR diffs, git diffs, or specified code sections. Maps every line to its screen/URL, data flow, and execution context like a debugger's step-through. | Serena MCP (recommended) |
| dnd | Browser drag-and-drop QA via coordinate-based pointer ops, plus video + drop+10-frame evidence for motion-sensitive bugs (DragOverlay rollback, ghost return). Knowledge-injection skill loaded by browser-using skills (e.g. ux-gap-detector) before any browser interaction — ref-based drag returns false success on dnd-kit and similar libraries. |
playwright-cli (required), ffmpeg (recommended for frame extraction) |
| electron-release | Guides Electron app release process including build, code signing, notarization, and GitHub Release with auto-update support. | — |
| explain | Deep, systematic explanation of code, concepts, and system behavior. Always operates at advanced level with introspection markers and validation. | Serena MCP (recommended), Context7 (recommended) |
| explain-diff-html | Explains a diff, branch, or PR as one self-contained interactive HTML page — deep background, core intuition with toy-data figures, a grouped code walkthrough, and a five-question multiple-choice quiz that checks the reader actually understood the change. Explores surrounding code for context; diagrams are HTML, never ASCII. | — |
| github-actions-pnpm-ci | Creates secure pnpm/Node GitHub Actions CI with SHA-pinned actions, pnpm store caching, frozen installs, lint/test/build/typecheck workflows, and Dependabot updates. | — |
| laststance-publish-skill | Publishes a stable skill to the laststance/skills GitHub registry for distribution via npx skills add. Updates README install commands, skills table, and usage examples in alphabetical order. |
— |
| learn-html | Answers a general "how does X work?" question as one self-contained interactive HTML explainer (background → intuition → mechanics → quiz). Every factual claim must be sourced — docs lookup, a live measurement, or a repo grep — and anything unverified is flagged inline rather than smoothed into confident prose. Output language follows the question's language. | Context7 MCP (recommended) |
| load | Load project context from Serena MCP memory for session initialization. Discovers memories, reads project overview, and validates context sufficiency. | Serena MCP (required) |
| locate-ui-from-code | Code → screen: locate UI with on-screen highlight overlay (ring + file badge), reach logic branches (debugger, if, useEffect, handlers) by executing the user operations that trigger them, and capture DOM dump + highlighted screenshots. Agent runs the reach recipe itself and leaves the browser open for DevTools. Tool-agnostic — cursor-ide-browser MCP (Cursor), playwright-cli (Codex/Claude Code), chrome-devtools MCP when available. |
playwright-cli (Codex/Claude Code), cursor-ide-browser MCP (Cursor), chrome-devtools MCP (recommended) |
| pr-tour | Live onboarding tour of newly implemented code. Runs the target app in a debug session — the vscode-debug-mcp bridge and playwright-cli attach to the same Chrome — pauses at curated breakpoints inside the new code while driving the real UI, narrates each stop in chat mapping UI moments to exact file:line, and writes a replayable tour artifact (before/after screenshots + deep-trace-extension replay table). | playwright-cli (required), Debug MCP Bridge (required), Serena MCP (recommended) |
| product-inspiration | Provides UI/feature implementation inspiration by researching top-tier apps. Implements all proposed patterns in _trials/ folder for hands-on evaluation. | Tavily MCP (recommended) |
| prop-drill | Trace React prop-drilling paths from origin definition to leaf consumers. Shows the original prop definition as a clickable code block, the full drilling route as a table, and a Mermaid flowchart. | Serena MCP (recommended), Context7 (recommended) |
| react-query-key-jump | Jump from a TanStack React Query queryKey string (e.g. getDrawing) to the useQuery / useInfiniteQuery hook line where that key is defined. Skips invalidateQueries usage sites. |
ripgrep (required) |
| rec | Record a web or Electron-renderer flow as an annotated video with playwright-cli — action callouts + chapter cards — then extract frames to confirm how it actually looks. For vague "record that part / 動作確認して録画" asks: real-time driver code paced for a human to watch (typing delay, pauses, chapter cards), not an E2E-runner replay. | playwright-cli (required), ffmpeg (required) |
| save | Save session context to Serena MCP memory for cross-session persistence. Analyzes accomplishments, persists learnings, and creates session checkpoints. | Serena MCP (required) |
| search | Iterative multi-tool research. Picks the best-fit tool (WebSearch, WebFetch, Exa, Perplexity, Tavily, Context7, DeepWiki) for the question type, then switches tool families across up to 3 passes until a citation-backed answer is reached. | Exa MCP (recommended), Perplexity MCP (recommended), Tavily MCP (recommended), Context7 (recommended) |
| source-grounded-research | Produces source-grounded research briefs with citations, contradiction handling, and query logs. Research-only — no implementation or speculative answers without sources. | Context7 (recommended), web search / MCP (recommended) |
| ts-pattern-refactor | Detect and refactor conditional code to ts-pattern's match().with().exhaustive(). Refactors JSX branching, chained ternaries, and discriminated-union dispatch — but deliberately leaves plain single-condition if-chains alone. Codifies syntactic-form × context judgment criteria. |
ts-pattern (required), Context7 (recommended), Serena MCP (recommended) |
| type-expand | Expands TypeScript type aliases into concrete, primitive-resolved shapes — unions, intersections, generics, conditional types, infer-based types, and common utility types as far as statically resolvable. Use when IDE hover only shows alias names. | tsx (recommended), TypeScript project tsconfig.json (required) |
| ux-gap-detector | Detects UI/UX quality gaps in authenticated SaaS web apps via playwright-cli. Crawls app interior, captures screenshots, scores across 4 dimensions (Typography & Spacing, Interactive States, Content Hierarchy, Loading & Error UX), and generates an actionable Markdown gap report. Optionally creates GitHub Issues. |
playwright-cli (required), Serena MCP (recommended) |
| video | Inspect video frame-by-frame and capture-then-verify UI motion. Extract frames from any clip with ffmpeg and read them as images; record interactions (Playwright, computer-use, iOS simulator) to verify animations and transitions that static screenshots and getComputedStyle cannot reveal. |
ffmpeg (required), Playwright (recommended for web/Electron renderer capture), iOS Simulator MCP (iOS capture), Computer Use MCP (native macOS chrome capture) |
| visual-lint | ESLint for rendered UI. Screenshots a running app via playwright-cli to detect wrapping, clipping, overlap, misalignment, and unreadable badge text. Checks colored badges against their backgrounds and explicit foreground requirements in light/dark themes. Baseline-free and read-only — reports cited findings without editing source or claiming unmeasured contrast ratios. |
playwright-cli (required) |
After installation, invoke skills as slash commands in your AI coding assistant:
/apollousa # Create PR, resolve CodeRabbit, merge, and clean up
/chrome-clean-install Chrome Canary # Clean-refresh a Chromium browser profile/cache
/ci-hardening # Port skills-desktop CI security baseline to this repo
/claude-code-plugin-troubleshoot # Debug Claude Code plugin issues
/code-trace # Trace code execution paths
/codebase-litter-audit # Find half-finished codebase litter
/coderabbit-resolver 17 # Process PR #17
/coderabbit-resolver --bulk # Process all open PRs
/colorful-type # Replace primitives with domain types
/component-hierarchy Button.tsx # ASCII tree from Page down to target component
/cookie # Import Chrome cookies into a playwright-cli session
/create-worktree feat/new-thing # Create git worktree at ../project-feat-new-thing
/deep-trace 42 # Trace PR #42 line-by-line
/dnd # Load drag-and-drop coordinate-based verification protocol
/electron-release # Electron release workflow
/explain src/auth/middleware.ts # Deep code explanation
/explain-diff-html 42 # Interactive HTML PR walkthrough + quiz
/github-actions-pnpm-ci # Add secure pnpm GitHub Actions CI
/laststance-publish-skill # Publish a stable skill to laststance/skills
/learn-html compositionend # Sourced HTML explainer: background → mechanics → quiz
/load # Load session context from Serena MCP
/locate-ui-from-code FolderHeader # Locate render target with screenshot + DOM dump
/locate-ui-from-code src/Foo.tsx:139 # Locate UI + reach debugger/effect/handler at pinned line
/pr-tour # Live debug-session tour of newly written code
/product-inspiration # Get UI/feature inspiration
/prop-drill orderData OrderTable # Trace prop-drilling path
/react-query-key-jump getDrawing # Jump to useQuery queryKey definition line
/rec あそこの部分 # Record a pointed-at flow, annotate, verify on frames
/save # Save session context to Serena MCP
/search what changed in React 19 # Iterative multi-tool research (Web + MCPs) until satisfied
/source-grounded-research React 19 # Cited research brief (no code changes)
/ts-pattern-refactor # Sweep codebase for ts-pattern refactor opportunities
/type-expand OrderItemSetting # Expand a TypeScript type alias to its concrete shape
/ux-gap-detector # Detect UX gaps in authenticated SaaS web app
/video clip.webm # Extract frames with ffmpeg and verify UI motion
/visual-lint # Check rendered UI defects and badge readability across themes
MIT