Latest Agent Skills

ALIAS for dag-orchestrator. The original DAG execution skill, now unified with orchestrator into dag-orchestrator. Use dag-orchestrator for the full HTDAG experience.

Guides creation of high-quality Agent Skills with domain expertise, anti-pattern detection, and progressive disclosure best practices. Activate on keywords: create skill, review skill, skill quality, skill best practices, skill anti-patterns, improve skill, skill audit. NOT for general coding advice, slash commands, MCP development, or non-skill Claude Code features.

Expert MCP (Model Context Protocol) server developer creating safe, performant, production-ready servers with proper security, error handling, and developer experience. Activate on 'create MCP', 'MCP server', 'build MCP', 'custom tool server', 'MCP development', 'Model Context Protocol'. NOT for using existing MCPs (just invoke them), general API development (use backend-architect), or skills/agents without external state (use skill-coach/agent-creator).

Executes DAG workflows with parallel wave processing, agent spawning, context isolation, permission enforcement, and full execution tracing. Use when running a planned DAG, managing concurrent agent execution, enforcing isolation boundaries, or tracing execution for debugging. Activate on "execute DAG", "run workflow", "spawn agents", "parallel execution", "execution trace", "agent isolation". NOT for planning DAGs (use dag-planner), validating outputs (use dag-quality), or matching skills (use dag-skills-matcher).

Validates agent outputs against schemas and quality criteria, scores confidence, detects hallucinations, monitors convergence, decides when to iterate, and synthesizes actionable feedback. Use when checking if a node's output is acceptable, scoring confidence, detecting fabricated content, deciding whether to re-execute, or generating improvement feedback. Activate on "validate output", "check quality", "confidence score", "hallucination check", "should we iterate", "improvement feedback". NOT for executing DAGs (use dag-runtime), planning DAGs (use dag-planner), or matching skills (use dag-skills-matcher).

Builds, validates, schedules, and dynamically modifies DAG execution graphs. Decomposes problems into nodes with dependencies, performs topological sorting, detects cycles, resolves conflicts, and schedules wave-based parallel execution. Use when designing a DAG structure, validating dependencies, planning execution order, or modifying a DAG at runtime. Activate on "build DAG", "plan workflow", "DAG dependencies", "topological sort", "schedule execution", "modify DAG", "replan". NOT for executing DAGs (use dag-runtime), validating outputs (use dag-quality), or matching skills to nodes (use dag-skills-matcher).

Build production-ready LLM applications, advanced RAG systems, and intelligent agents. Implements vector search, multimodal AI, agent orchestration, and enterprise AI integrations. Use PROACTIVELY for LLM features, chatbots, AI agents, or AI-powered applications.

This skill should be used when a SKILL.md file needs compression, deduplication, or token reduction. It provides an embedding-based compression pipeline that detects and removes redundant chunks within SKILL.md files using local embeddings (all-MiniLM-L6-v2). Two-pass approach: (1) free intra-skill deduplication via cosine similarity clustering, (2) optional LLM-judged graded eval to detect pretraining overlap. Typical result: 25-46% token reduction with zero quality loss. This skill is not intended for editing skill content, creating new skills, routing optimization, or cross-skill deduplication.

Progressive Web App development with Service Workers, offline support, and app-like behavior. Use for caching strategies, install prompts, push notifications, background sync. Activate on "PWA", "Service Worker", "offline", "install prompt", "beforeinstallprompt", "manifest.json", "workbox", "cache-first". NOT for native app development (use React Native), general web performance (use performance docs), or server-side rendering.

Selects the optimal LLM model and provider for each task based on complexity, cost budget, and capability requirements. Routes cheap tasks to Haiku/GPT-4o-mini and complex tasks to Sonnet/Opus/o1. Use when deciding which model to call, optimizing LLM costs, or building multi-model agent systems. Activate on "which model", "model selection", "route to model", "LLM cost", "model routing", "cheap vs expensive model". NOT for prompt engineering (use prompt-engineer), model fine-tuning, or training custom models.

Professional CV and resume builder transforming career narratives into ATS-optimized, multi-format resumes. Integrates with career-biographer for data and competitive-cartographer for positioning. Generates PDF, DOCX, LaTeX, JSON Resume, HTML, and Markdown. Activate on 'resume', 'CV', 'ATS optimization', 'job application'. NOT for cover letters, portfolio websites (use web-design-expert), LinkedIn optimization, or interview preparation.

Predicts the highest-impact next action for your project by running a 5-agent meta-DAG pipeline. Gathers project signals automatically (git, recent files, port-daddy, CLAUDE.md), then runs sensemaker → decomposer → skill-selector + premortem → synthesizer. When execution is approved, convert each predicted node into a skillful node prompt using skillful-node-prompt + skillful-subagent-creator, prefer live WinDAGs visualization backed by POST /api/execute and /ws/execution/:id, and fall back to ASCII only when live visualization is unavailable. Activate on: "what should I do", "what's next", "next move", "/next-move", "where should I focus", "what's the highest impact thing right now". NOT for: creating skills, debugging one specific bug, or promising topology-specific runtime behavior the current server cannot execute.

Human interface agent that translates ecosystem activity into clear, actionable communication. Creates status briefings, decision requests, celebration reports, concern alerts, and opportunity summaries. Use for 'status update', 'brief me', 'what's happening', 'summarize progress', or when complex multi-agent work needs human-readable reporting.

Philosophy of science methodology examining how research programs evolve through proofs and refutations

Operations, debugging, and optimization for DAG workflows. Performs root cause analysis on failures, profiles execution performance, aggregates results from parallel branches, bridges context between nodes, and learns patterns from execution history. Activate on "DAG failed", "why did it fail", "root cause", "performance profile", "aggregate results", "merge branches", "execution patterns", "optimize DAG". NOT for planning DAGs (use dag-planner), executing DAGs (use dag-runtime), or validating outputs (use dag-quality).

Cloudflare Pages and Workers deployment checklist for wrangler deploys, secret audits, D1 migrations, Worker-compatible SDK checks, and browser-auth verification. Use when shipping Cloudflare changes to production. NOT for general CI/CD architecture, non-Cloudflare deployments, or post-launch UX QA.

Read-only audit of a codemod or bulk mechanical transform for behavior-preservation — compiles/lints is not same-behavior. Use when "did this codemod break anything", "audit this bulk refactor", or before merging a mass find-replace. Diff quality → audit-code-review. SQL → plan-data-integrity.

Drive an entire repository to a fully green baseline — typecheck, lint, tests, and build all passing from a clean checkout — when the user has explicitly authorized fixing. Use when "make the repo green", "get CI passing", "fix all the failing tests". Ratchet integrity → audit-gate-logic.

Generate clear, descriptive commit messages following conventional commits format. Use when committing code, writing commit messages, or when the user asks for help with git commits.

OWASP Top 10 + Supabase-first hardening burndown. Use when "security audit plan", "OWASP audit", "hardening plan", or "security burndown". App-layer auth flows → audit-auth-flows. Table RLS → plan-rls-audit. Key rotation → plan-secrets-audit. App LLM attacks → audit-llm-security.

Audit a codebase for silent failures, swallowed exceptions, and observability gaps across Sentry and Langfuse, then produce a phased fix plan. Use when the user says "errors aren't showing in Sentry", "things fail silently", "empty catch blocks", "add error handling", "check my Langfuse tracing", or is hardening.

Audit a project for destructive-operation and migration safety gaps, then produce a phased safeguard plan. Use when "is my migration safe", "could I lose data", "my agent might delete prod", or "safe schema changes". Restore drills and RPO/RTO belong to plan-backup-dr. Source transforms → audit-codemod-safety.

Close the post-launch improvement loop for any shipped app. Use when asked to "improve the app after launch", "fix the top issues", "post-launch polish", "what should I fix next", "production issues", "iterate on feedback", "post-release improvements", "what is broken in prod", "ship a polish pass", or "make it better.

Build landing pages, portfolios, and marketing sites that don't look AI-generated. Use when asked for "landing page", "portfolio", "marketing site", "anti-slop", "Awwwards-style", "premium frontend", or when design needs a strong point of view.

Design mobile-first UIs: touch targets, safe areas, gestures, then enhance up. Use when building for small screens, touch, swipe, PWA, or tablet. Linearized desktop / responsive audit → audit-responsive.

Close an approved plan with zero plan-related deferrals: implement every unfinished item, absorb every connected out-of-scope/follow-up/nice-to-have item parked by prior runs, fix. Use when a plan was marked done with work deferred, or the user says "complete everything", "don't defer", "fix out of scope too", "finish.

Automated WCAG 2.2 accessibility audit using playwright-cli to crawl every page, inject axe-core via eval, test keyboard navigation, check color contrast, ARIA labels, and heading order. Use when "audit accessibility", "check a11y", "WCAG audit", "check keyboard nav", or "test screen reader".

Split current work into small reviewable PRs. Use when the user asks to split a chat, set of changes, branch, or PR.

Guide users through creating effective Agent Skills for Cursor. Use when user wants to create, write, update, or debug a skill, or asks about SKILL.md format, skill structure, ~/.cursor/skills/, or skill best practices.

Repo housekeeping: sync READMEs to match current architecture, remove dead files (logs, screenshots, deprecated code, build artifacts), update npm/pip/cargo dependencies to latest. Use when asked to: "housekeep", "clean up repo", "update README", "update dependencies", "fix vulnerabilities", "remove dead code", "tidy.

User-story-driven test coverage audit and plan — no test writing in this pass. Use when "test coverage plan", "coverage audit", "traceability matrix", "fake-green tests", "uncovered user stories", "plan tests for critical flows", or "whats not tested". Mutation score / assertion theater → test-mutation.

Audit the working tree and git history for exposed credentials and mis-scoped keys, then a rotate-vs-relocate plan. Use when "hardcoded secrets", "did I commit a key", "secret scan", "is my .env safe", or "rotate keys". Do not fire for "RLS audit" or generic "security burndown".

Audit a site for answer-engine / GEO citation readiness (ChatGPT, Perplexity, AI Overviews), then a phased plan. Use when "AEO", "GEO", "show up in ChatGPT/Perplexity", "AI search visibility", "llms.txt", or "am I blocking AI crawlers". Classic SEO meta/sitemap → enhance-web-seo.

Create or update Cursor agent skills (SKILL.md). Use when authoring a new skill, updating an existing skill, or asking about SKILL.md structure and best practices.

Build interactive, accessible charts, graphs, and data dashboards using Recharts, D3, or Victory. Use when the user says "chart", "graph", "data visualization", "analytics dashboard", "metrics display", "D3", "Recharts", "time-series", or "data display".

Read-only OWASP LLM Top 10 audit of app-facing AI features: prompt injection, data leak, supply chain, poisoning, unsafe output, excessive agency, system-prompt leak, RAG/embedding risks, misinformation, unbounded consumption. Use when "audit LLM security", "prompt injection", "jailbreak my chatbot", "is my AI safe".

Run a PDCA quality audit on LLM/AI features: traces, prompts, costs, evals, grounding, hallucination. Use for "audit LLM quality", "check Langfuse", "audit prompts", "check AI quality", "audit AI costs", "check traces". Jailbreak/OWASP LLM → audit-llm-security. Token caps → plan-llm-cost-guardrails.

Detect and fix repo-wide anti-patterns and consistency drift (naming, organisation, repeated smells). Use when "code smell", "anti-pattern", "technical debt", or "standardize the codebase". This PR/diff review → audit-code-review.

Create Cursor rules for persistent AI guidance. Use when user wants to create a rule, add coding standards, project conventions, file-specific patterns, or mentions ".cursor/rules", "AGENTS.md", "cursor rule", "coding standard", "convention", or "AI guidance".

Create Cursor hooks. Use to create a hook, write hooks.json, add hook scripts, or automate behavior around agent events.

Stop vibe-coding with a spec → plan → TDD loop before writing a line. Use for any non-trivial feature, refactor, or bug fix on web, React Native, or Capacitor. Use when asked to "build", "implement", "add a feature", "this keeps breaking", "do it properly", or when LLM output keeps being wrong.

Guide for refactoring code to improve quality without changing behavior. Use when refactoring, cleaning up code, reducing duplication, improving readability, or restructuring code. Integrates Firecrawl for researching modern patterns before refactoring, and codebase-aware dependency analysis to avoid breaking changes.

Exhaustive audit for stubs, dead buttons, fake components, unwired handlers, and dead links — then a wiring plan, no implementation. Use when "find dead buttons", "stub checker", "fake components", "unwired handlers", or "dead links". Live QA → test-qa.

Full end-to-end Mushi Mushi integration smoke test: bug capture → AI triage → story mapping → TDD test generation → approval → execution → PDCA cycle. Use when "test mushi integration", "verify full pipeline", "mushi e2e check", "does mushi work end-to-end", "smoke test mushi", or after deploying changes.

Add 3D and scroll-driven motion to an existing website or web app — Three.js / React Three Fiber for the scene, GSAP ScrollTrigger for scroll-driven choreography, and Motion /. Use when the user asks to "add 3D", "add a WebGL hero", "make it cinematic", "scroll-driven 3D", "three.js scene", "React Three Fiber", "GSAP.

Build or upgrade web forms to production quality: accessible structure, schema-driven validation, client↔server parity. Use when "improve this form", "form validation", "accessible form", "multi-step form", "form error handling", or "the form UX is bad".

Cross-page UX audit for user stories, task completion, and information architecture — the layer audit-ux (per-page heuristics) skips. Use when "audit user flows", "IA audit", "can users find X", "navigation audit", or "funnel drop-off". Full DS burndown → plan-uiux-unification.

Audit and optimize application runtime performance (Core Web Vitals, slow code, load time). Use when "slow page", "LCP/INP/CLS", or "optimize performance". JS payload → audit-bundle-size. Concurrent breaking point → test-load. Timeouts/retries → audit-resilience.

Review this PR or diff for quality, security, and maintainability. Use when reviewing pull requests, examining a named change set, or the user asks for a code review of current work. Repo-wide anti-patterns → audit-code-quality. Bulk transform semantics → audit-codemod-safety.

Analyse and shrink JavaScript bundle size for any web app. Use when asked to "reduce bundle size", "analyse bundle", "tree shaking", "lazy loading", "code splitting", "slow initial load", "large JS", "chunk size", "build performance", "LCP caused by JS", "why is the bundle so big", or "first load JS too large".

Configure a custom status line in the CLI. Use when the user mentions status line, statusline, statusLine, CLI status bar, prompt footer customization, or wants to add session context above the prompt.

First-contact orientation for an unfamiliar codebase. Use when "I'm new to this repo", "orient me", "explain this codebase", "what does this do?", "onboard me", "first day on this project", or "catch me up on the codebase".

Exploratory ("monkey") QA of a live app as GUEST then LOGGED-IN, then diff. Headed playwright-cli, junk inputs, nav abuse. Use when "monkey test the app", "exploratory QA", "wander like a confused user", or "guest vs logged in". Story CRUD → test-qa. Hostile sweep → test-red-team. This-diff PDCA → test-playwright.

Polish an existing React Native screen to feel intentional, native, and human-crafted. Use for "this screen looks off", "feels clunky on iOS", "Android version looks wrong", "jank when scrolling", "button is unreachable", or any RN-specific UX polish pass.

Scaffold and implement Model Context Protocol (MCP) servers that expose external services, APIs, and data sources as typed tools and resources for LLM agents. Use when the user says "build an MCP server", "give Claude access to X", "create an MCP tool", "expose my API to an agent", or "AI agent integration".

Apply-now consolidation of a drifted design system into one token/component SSOT. Use after an approved plan-uiux-unification pass, or when "clean up the design system now", "migrate to one button", or "resolve token conflicts". Plan-only burndown → plan-uiux-unification.

Implement real-time features using WebSockets, Supabase Realtime, Server-Sent Events, and live data. Use when user wants "real-time", "live updates", "WebSocket", "notifications", "chat", "collaborative", "presence", "live data", or "instant sync".

Apply modern backend patterns — auth middleware, caching strategies, background queues, rate limiting, and serverless/edge function design — across stacks (examples use Next.js. Use when the user says "backend architecture", "queue jobs", "caching layer", "rate limiting", "server actions", "edge function".

Audit-and-fix linearized mobile layouts at every breakpoint — desktop is not a wide phone. Use when "responsive audit", desktop looks stacked/stretched, no max-width, breakpoint gaps, or 1440 looks like a phone. Distinct from design-mobile-first (touch/mobile-up) and audit-ux-journeys (cross-page IA).

Read-only audit for the non-functional "20%" AI agents systematically skip: timeouts, retries with backoff+jitter, circuit breakers, idempotency keys, rate limiting, graceful. Use when "is this production-ready", "resilience audit", "will this survive real traffic", "audit retries/timeouts/idempotency", "reliability.