agent-skills

Operate the official grafana/mcp-grafana server from an agent. Use when the conversation has Grafana MCP tools available and you need to query Prometheus/Loki/Pyroscope, search or edit dashboards, manage alert rules, or generate Grafana deeplinks. Covers tool selection, parameter conventions (datasource UID, time ranges, limits, query types), and context-saving patterns for dashboard work.

Analyzes an existing ISUCON-style application, designs an optimal infrastructure and architecture redesign that respects every ISUCON rule, and emits the proposal as a self-contained Agent Skill that another Claude session can read and execute. Use when starting an ISUCON contest or practice, drafting a contest-day architecture plan before touching code, or when the user asks to "design the architecture", "plan the topology", or "produce an implementation skill" for an ISUCON problem.

Decomposes the apply-skill produced by designing-isucon-architecture into atomic tasks, materializes one Claude Code subagent per task, and emits a dispatcher slash command that fires the subagents in dependency order (parallel where the graph allows). Use when an applying-isucon-arch-<slug> skill exists and the team wants the architecture rollout to run as orchestrated subagent jobs rather than a single linear human-driven session, or when the user asks to "create subagents for this ISUCON design", "wire the ISUCON apply-skill into Claude Code agents", or "split the rollout into parallel agent tasks".

Runs Visual Regression Testing (VRT) locally to prevent disqualification in Web Speed Hackathon. Captures screenshots, compares against baselines, updates snapshots, and validates visual integrity after performance optimizations. Use when optimizing WSH apps, running VRT checks, updating VRT baselines, or investigating VRT failures.

Analyzes a Web Speed Hackathon project, researches each improvement topic in depth, stores research as Agent Skills, and generates executable Claude Code commands for each optimization. Use when starting WSH optimization, creating an optimization roadmap, or generating executable improvement commands.

Ports an ISUCON-style Rust reference implementation to Go (Echo) using parallel subagents. Use when the contest provides a Rust reference but the team wants to compete in Go, or when the user asks to migrate a Rust web app to Go Echo endpoint by endpoint. Preserves API URI, request/response shape, frontend static files, the isuwari reboot/retest daemon, and the isuadmin account exactly as distributed (changing them = disqualification).

Optimizes web applications for ISUCON (Iikanji ni Speed Up Contest), a Japanese 8-hour performance tuning competition. Use when participating in ISUCON, practicing past problems (private-isu, isucon-workshop), or aggressively tuning a Go/Node.js/Ruby/PHP web app with MySQL behind nginx. Covers regulation, measurement tools (alp, pt-query-digest, pprof), N+1 elimination, indexing, caching, multi-server topology, and reboot-test pitfalls.

Optimizes deliberately slow web applications for maximum Lighthouse scores in Web Speed Hackathon (CyberAgent). Use when participating in WSH or performing aggressive frontend performance optimization on React/Node.js apps with SQLite backends. Covers bundle reduction, image optimization, Core Web Vitals, server tuning, and known competition traps.

Run commands directly against ISUCON contest hosts (isu1/isu2/isu3) over plain SSH. Use when the user wants to ssh into a contest box and execute commands — checking app/nginx/mysql status, tailing journalctl, restarting services, opening a MySQL shell, running kataribe or pt-query-digest, profiling with pprof through an SSH tunnel, transferring files, or running anything ad-hoc that doesn't have (or doesn't need) a Makefile target. Covers `~/.ssh/config` + agent-forwarding setup, host inventory, server-side paths and service names, common one-liners by category, and SSH patterns (heredoc, port-forward, multi-host, scp/rsync).

Deploys ISUCON contest code and configs to competition servers using the isucon-ansible layout (Ansible playbook for provisioning + Makefile-over-SSH for the per-benchmark loop). The two deploy commands are `make bench` (regular deploy, with instrumentation ON) and `make maji` (final-run deploy, with instrumentation OFF). Use when running benchmarks, deploying app/nginx/MySQL config changes, reassigning roles between servers, or when working in a repo that uses mazrean/isucon-ansible (server.yaml, hosts inventory, .make.env, remote/Makefile).

Creates implementation tasks as Claude Code custom slash commands with dependency ordering and atomic scope. Use when breaking down features into executable task commands, planning implementation order, defining task dependencies, or when user mentions task breakdown, implementation plan, or work decomposition for spec-driven development.

Creates feature specifications (PRD) as Agent Skills with progressive disclosure. Use when writing product requirements, feature specs, user stories, acceptance criteria, or when starting spec-driven development for a new feature. The PRD is output as a skill directory with SKILL.md and reference files.

Creates agent-optimized technical design documents backed by deep research of every technical component. Each component (library, framework, protocol, service) is investigated via web search / official docs and distilled into its own Agent Skill under skills/tech-{component}/, so future implementation sessions auto-load the relevant knowledge. Use when writing technical designs, architecture docs, defining system components, or making technology choices for spec-driven development.

Releases Zig CLI applications using GoReleaser v2.5+. Use when setting up automated releases for Zig projects, configuring .goreleaser.yaml for Zig builds, creating GitHub Actions CI/CD for Zig releases, or publishing Zig binaries via Homebrew, Docker, or Linux packages.

Writes Zig code for sharing TCP/UDP sockets between unrelated processes using SO_REUSEPORT. Use when implementing multi-process socket sharing, zero-downtime restarts, load balancing across processes, or when user mentions SO_REUSEPORT, reuseport, socket sharing, or graceful restart in a Zig project.

Integrates UnoCSS atomic CSS engine with Go templ templates for static sites. Use when setting up UnoCSS in templ projects, configuring uno.config.ts for .templ files, using utility classes in templ components, or building static sites with Go and atomic CSS.

Creates project constitution files (CLAUDE.md/AGENTS.md) that serve as always-loaded context for coding agents. Use when setting up a new project for spec-driven development, configuring agent instructions, writing CLAUDE.md or AGENTS.md, or establishing project-wide coding standards and constraints.

Writes CLI tools in Zig using modern 0.15+ syntax and best practices. Use when creating command-line applications in Zig, parsing CLI arguments, setting up Zig project structure with build.zig, or working with Zig's I/O and error handling for CLI programs.

Builds fast, lightweight Web Components using Lit library. Use when creating Lit components, defining reactive properties, handling lifecycle, using directives, styling with Shadow DOM, implementing Context API, Task for async, or SSR with Lit.

Creates static websites with RSS/Atom/JSON Feed support for content syndication. Use when building static sites with feeds, implementing RSS in Hugo/Jekyll/custom generators, adding feed autodiscovery, or when user mentions RSS, Atom, feeds, or content syndication.

Builds type-safe HTML components in Go using templ templating language. Use when creating templ components, defining component parameters, handling children, conditional rendering, loops, CSS styling, JavaScript integration, or setting up templ with HTTP servers.

Creates clean, atomic commits following Conventional Commits specification. Use when committing code changes, writing commit messages, or when user mentions commit, conventional commits, or atomic commits.

Creates well-structured Agent Skills following best practices. Use when building new skills for Claude Code, designing skill directory structures, writing SKILL.md files, or improving existing skills with progressive disclosure patterns.

Designs comprehensive test cases using systematic techniques (boundary value analysis, equivalence partitioning, mutation testing, property-based testing). Use when writing tests, improving test coverage, finding edge cases, or when user mentions test cases, testing, or test design.

Manages Go development tools using the Go 1.24+ tool directive in go.mod. Use when adding linters, generators, or dev tools to Go projects, running project-specific tools, integrating tools with go:generate, or migrating from tools.go patterns.