cskwork

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

Raw mechanical interfaces fusing Swiss typographic print with military terminal aesthetics. Rigid grids, extreme type scale contrast, utilitarian color, analog degradation effects. For data-heavy dashboards, portfolios, or editorial sites that need to feel like declassified blueprints.

Claude Code의 statusline HUD 플러그인. context 사용량, tool 활동, agent 추적, todo 진행상황을 실시간으로 표시한다. statusline 설정, HUD 커스터마이징, 또는 context 모니터링이 필요할 때 사용한다.

Use when a user asks how to install plugins from this marketplace, needs onboarding to the Claude Code plugin system, or encounters plugin installation errors like "marketplace not found" or "plugin not registered"

한국 시장 대상 프론트엔드 프로젝트의 공통 로컬라이제이션 오버라이드. Pretendard 폰트, word-break keep-all, 한국어 콘텐츠 패턴, standalone HTML 출력. taste-skill 계열 디자인 skill과 함께 적용.

Clean editorial-style interfaces. Warm monochrome palette, typographic contrast, flat bento grids, muted pastels. No gradients, no heavy shadows.

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

Upgrades existing websites and apps to premium quality. Audits current design, identifies generic AI patterns, and applies high-end design standards without breaking functionality. Works with any CSS framework or vanilla CSS.

Implement code using sonnet model with full main context access

Start the multi-AI pipeline with a given request. Guides through plan -> review -> implement -> review workflow.

Final code/plan review using Codex CLI. Use as the last review step after sonnet.

Fast code/plan review for quality, security, and tests. Use for quick reviews before deeper analysis.

Implement code using sonnet model with full main context access

Start the multi-AI pipeline with a given request. Cleans up old task files and guides through plan → review → implement → review workflow.

Final code/plan review using Codex. Use as the last review step after sonnet and opus.

Deep code/plan review for architecture, subtle bugs, and test quality. Use after sonnet review for thorough analysis.

Use when starting a non-trivial backend task in a legacy or multi-service codebase — especially when you've never touched the affected area, the blast radius is unclear, or a quick fix would be tempting but risky

Use when a change has been deployed to a dev/stg/audit/prod project environment and browser-level E2E proof against the deployed URL is needed — typically after /verify passes and the deploy lands, for post-deploy smoke checks, or standalone when a specific deployed flow needs ad-hoc browser verification. Not for localhost — /verify owns that.

Use when a user-reported or ticketed browser issue must be reproduced in an actual Playwright browser before coding, especially when exact URL, account/role, data state, screenshots, network calls, and as-is/to-be comparison evidence are needed.

Use when code changes have been made to backend HTTP endpoints and HTTP-level proof against a running service is needed — typically after /work completes, or standalone when you have an endpoint plus at least one payload to probe ad-hoc

Use when an approved exploration report exists at <artifact-root>/.backend/<YYYYMM>/<slug>/explore.md, including service-local .backend folders, and it is time to implement the plan

Use when bootstrapping Symphony into an existing monorepo (multi-service Git repo) so each ticket gets an isolated git worktree workspace, the upstream 7-stage prompts are installed, and Claude Code permissions are wired up bidirectionally. Triggers on "set up symphony in my monorepo", "bootstrap symphony workspaces", "WORKFLOW.md per service", "worktree hooks", "symphony monorepo".

Use for Symphony OneShot runs that turn one prompt into ordered Kanban tickets with shared vault state, verification gates, QA evidence, and delivery proof.

Use for Symphony Kanban operations including tickets, TUI/service runs, workflow prompts, delegation, and dispatch or worker failure triage.

Use for "fix this bug", "add/build/plan/spec this feature", "prototype this", "QA / verify", "code review", "improve architecture", "teach/explain", "learn codebase", "make skill", or "eval a harness".

Create professional architecture, workflow, sequence, data-flow, and lifecycle/state diagrams as standalone HTML files with SVG graphics, a built-in dark/light theme toggle, and one-click export to PNG / JPEG / WebP / SVG. Accepts plain-language descriptions or pasted Mermaid code (flowchart, sequenceDiagram, stateDiagram) and lays the diagram out from scratch in archify style. Use when the user asks for system architecture diagrams, infrastructure diagrams, cloud architecture visualizations, security diagrams, network topology, technical workflows, approval flows, runbooks, CI/CD flows, process diagrams, API call sequences, request lifecycles, data pipelines, ETL/ELT maps, PII boundaries, data lineage, state machines, lifecycle diagrams, status transitions, or asks to convert/beautify a Mermaid diagram.

Empirically measure two things for any task or domain: (1) does loading a SKILL.md actually change behavior — with_skill vs without_skill (baseline) — and (2) which CLI agent harness does the task best — claude vs codex vs gemini vs agy vs an OpenAI-compatible API. Runs the same prompt across the matrix in fresh, isolated contexts (subagents or separate CLI processes), a judge grades every output, repeated over trials, and you get a skill-lift table plus a harness leaderboard. Use to validate a new skill before shipping, catch skills that do nothing (or hurt), compare CLIs on a task you give, or pick the best harness for a domain. Triggers: evaluate skill, test skill, does my skill work, skill A/B, measure skill lift, compare CLIs, which agent is best, claude vs codex vs gemini, harness eval, benchmark agents on a task.

Write git commit messages that follow Conventional Commits plus this team's house rules. Use when writing or rewriting a commit message.

Use when learning a codebase another agent (human or AI) has built, reviewing an AI-generated PR against a spec, onboarding to a single feature mid-stream, planning a change to behavior whose blast radius you do not yet know, or diagnosing suspected spec-versus-code drift. Applies whether intent docs and implementation live in the same repo or in separate planning + code repos.

Use when the user wants to be quizzed on a codebase to verify their own understanding of logical flow, domain rules, invariants, or design decisions — instead of being told. Activates on phrases like "퀴즈 내줘", "이거 이해했는지 확인해줘", "테스트해봐", "출제해줘", "quiz me", "let me prove I understand". Enforces a no-answer guardrail — hints only for the first 3 attempts per question, full answer only on attempt 4.

Archive (not delete) Claude Code and Codex skills that are zero-use or stale (>90 days unused) based on a generated skill-stats.html report. Renames the SKILL.md inside each archived directory so agents stop discovering them, while keeping all content recoverable. Triggers — "clean up skills", "archive unused skills", "스킬 정리", "안 쓰는 스킬 정리", "skill archive".

Generate an HTML dashboard of skill usage across Claude Code and Codex (global + project .claude/skills, ~/.codex/skills, ~/.codex/prompts). Counts invocations, last-used date, days-since-use, session reach, and on-disk size so you can decide which skills to keep, archive, or delete. Triggers — "skill usage stats", "which skills do I actually use", "clean up skills", "audit my skills", "스킬 사용 통계", "안 쓰는 스킬", "스킬 정리".

Single Symphony operator router for Kanban tickets, service/TUI runs, workflow prompts, delegation, production-ready app delivery planning, OneShot runs, monorepo bootstrap, and worker failure triage.

Use whenever the user mentions "multica", asks to create/move/comment/assign a board ticket, dispatches an AI coding tool (Claude Code, Codex, Cursor, Gemini, Copilot, ...) to an issue, manages workspaces/projects/agents/skills/squads/daemon/runtimes/autopilots, sets up the CLI or a PAT, onboards a new machine, or runs a multi-phase coding pipeline on a Multica board. Covers the full `multica` CLI plus onboarding and the explore→work→review→qa→learn workflow.

Local text-to-speech via the supertonic-tts CLI. Use when the user asks to "speak", "narrate", "say", "read aloud", "make audio from text", "TTS", "음성 생성", "내레이션", "읽어줘". Generates 44.1 kHz WAV from EN / KO / JA (and 29 other language tags). No API key, fully offline after the first asset download. Auto-plays by default.

Convert PowerPoint (.pptx) presentations to standalone HTML format with FULL style, position, and formatting preservation. Accurately replicates slides with exact fonts, colors, shapes, backgrounds, layouts, hyperlinks, videos, audio, and tables. Use for web-friendly presentations that maintain visual fidelity and interactivity.