my-claude-plugins

Generate or edit bitmap images from Claude Code by delegating to Codex CLI image generation, without managing OpenAI API keys. Use when the user asks to generate or edit a raster image, or when an active task's spec explicitly designates codex-image as the image path (e.g. deck builds); never research raw codex invocation instead of loading this skill. When the grounding or scope is ambiguous, confirm before generating — image generation has cost and side effects.

Use when the user asks an in-depth question about a GitHub repository's internals, architecture, or implementation details — anything that benefits from AI-summarized repo documentation rather than raw code reading. Triggers on phrases like "how does <repo> handle X", "explain the architecture of <repo>", "what's inside <owner/repo>", or "compare <repoA> and <repoB>". Wraps the DeepWiki MCP (`mcp__deepwiki__read_wiki_structure` / `read_wiki_contents` / `ask_question`).

Onboard a full Playwright E2E test harness in the current project — verify/install Playwright, generate the official planner/generator/healer agents via npx playwright init-agents --loop=claude, scaffold auth separation (storageState + setup project), network route mocking, an E2E operating SSOT doc, and a gated GitHub Actions CI workflow with trace artifacts and PR-failure comments. Use when the user asks to set up E2E, add Playwright AI agents, bootstrap end-to-end testing, or wire E2E into CI. Never overwrites an existing playwright.config (merge proposal + backup). Degrades gracefully when Playwright is absent. Run from the user's project root, not this marketplace repo.

Use to scaffold the LLM-Wiki knowledge system (`.llmwiki/` insight + wiki + raw, plus spec) in a new repo that doesn't have it yet. Copies templates from the plugin's bundled assets, prompts for project pitch and 1-3 first domains, writes a slim CLAUDE.md.

Use after producing a new audit md, after merging a PR with non-obvious findings, or when a debugging session uncovers lore worth saving. Pulls the finding into the wiki with reversible diff-log + multi-page cross-update. Universal — works in any repo with a `.llmwiki/wiki/` or legacy `.claude/wiki/`.

Use when the user asks to audit wiki health, or periodically (manual trigger) to catch the 4 wiki-rot failure modes — identity duplication, level flattening, monotonic relationships, and staleness. Universal — works in any repo with a `.llmwiki/wiki/` or legacy `.claude/wiki/`.

Use to migrate or upgrade an existing wiki to the v2 `.llmwiki/` layout — triggers on "migrate wiki", "upgrade wiki to v2", "consolidate .claude/wiki and .codex/wiki", "move wiki to .llmwiki". Detects legacy `.claude/wiki/` and forked `.codex/wiki/`, consolidates them under the neutral `.llmwiki/` root, and adds v2 frontmatter. Universal — works in any repo with a legacy wiki.

Use as the verification gate BEFORE acting on remembered global/project guidance, provider quirks, design rationale, debugging stories, or module maps that aren't code invariants — check the wiki MOC first so you rely on the dated, sourced page instead of stale memory. Universal — works in any repo with a `.llmwiki/wiki/` or legacy `.claude/wiki/`.

Design and implement PyTorch GPU parallel processing pipelines for maximum throughput. Use when scaling workloads across multiple GPUs (ProcessPool, CUDA_VISIBLE_DEVICES isolation), optimizing single GPU utilization (CUDA Streams, async inference, model batching), or building I/O + compute pipelines (ThreadPool for loading, batch inference). Triggers on "multi-GPU", "GPU parallel", "batch inference", "CUDA isolation", "GPU utilization", "ProcessPool GPU", "PyTorch multi-GPU".

Build a Tally questionnaire/survey form from a checklist markdown — parse the md, run a copy-voice + humanize pass, build blocks with theme presets, section dividers, paragraph-split intros, per-question choices with required + checkbox (multi-select), short-answer inputs (text/number/email/phone/link), native scheduling (matrix grid, date, time), and form images (logo/cover/inline, URL-hosted) plus redirect-on-completion, then create or idempotently publish via the Tally API and return the share URL. Reusable per project or client. Use when the user wants to '설문 폼 만들어', 'Tally 폼 만들어', 'questionnaire', '체크리스트를 폼으로', '상담 신청 폼', '일정 조율 설문', '폼에 로고/이미지', '필수/복수선택/단답 문항', 'dev survey form', or 'lecture consultation form'.

Pre-flight CodeRabbit + Codex review state on the current PR, autonomously judge each finding (apply / defer / skip with reasoning), commit, push, and loop until clean. Use when the user types /github-dev:cr-fix, says "auto-fix the review", "process CodeRabbit feedback", or "loop until clean". v2 adds Step 5 pre-flight detection (skip wait when reviews already arrived) and removes the per-finding AskUserQuestion gate (LLM decides apply/defer/skip from code + severity, surfacing reasoning in the final report). Still handles PR-bot rate-limits with auto-fallback to local CodeRabbit CLI or Codex-only, and supports --auto-merge with branch-protection gating.

This skill should be used when the user asks to run a project setup diagnostic on an EXISTING repository — phrases like '프로젝트 진단', '셋업 점검', '하네스 배선 확인', '이 repo 설정 제대로 됐는지 확인', 'check my project wiring', 'is this repo wired up', 'diagnose my project setup', or an explicit /project-init:wiring invocation. Detects 14 axes of agent-harness configuration, including four that check whether config takes EFFECT rather than merely exists: git core.hooksPath, an @import that defeats .claude/rules paths: scoping, MCP servers registered twice so one copy is silently discarded, and the Codex AGENTS.md byte budget. Verdicts are FAIL / WARN / ASK / INFO / SKIP / OK — an ASK is a decision nobody made yet, asked once and persisted to .claude/state/wiring.json. Read-only until approved. Complements /project-init:new (empty dirs only). Not for mem0 store diagnostics (/mem0-ops:doctor) or wiki-content health (/llm-wiki:lint-wiki).

yeong 스타일 강의·제안 덱 작성 규약 — ppt-master 빌드 엔진 위에 얹는 작성 레이어(엔진 자체가 아님). 덱 유형·md 소스 규약·작성 원칙 16종·밀도 리듬(중간 강화 기본)·역할 기반 색 팔레트·codex-image vs SVG 경계·앱 UI 실물 강제·ppt-master 레버 조합 차별화·빌드 후 스토리 흐름 review·공식 로고 fetch·윤문·렌더 QA를 한 세션·어느 repo에서든 일관 적용. ppt-master/slidev로 그냥 'PPT/슬라이드 만들기'와 달리, yeong 특유의 담백한 명사구 제목·anti-slop 본문 + 표지/목차/전환 pop 무드·일상 비유 보조선이 필요할 때. 트리거 — 'yeong 스타일 덱', '강의 덱/실습 덱 만들어줘', '제안서 슬라이드 만들어줘', 'ppt-master로 강의자료', '발표자료 yeong 스타일로', 'yeong style lecture/proposal deck', even when this skill is not named. 일반 'PPT 만들어줘'는 ppt-master, 'slidev 슬라이드'는 slidev로 — 이 스킬은 yeong 규약이 필요할 때만.

yeong 강의 덱 전용 운영 규약 — ppt-yeong-style(작성 규약) 위에 얹는 강의 특화 레이어. 40~50장급 구성·실습 handouts 생성 규약(자료-지시문 정합 검증)·실습 프롬프트 카드·placeholder→실캡처 스크린샷 슬롯 운영·리넘버링 파이프라인(4중 동기화)·표 행 수 변경 재배치·강의 후 전사 회고 루프·강사 노트 운영 규약([시연 필수]·수치 고정 대본·주차장 멘트)을 다룬다. 트리거 — "강의 덱 만들어줘/보강해줘", "실습 자료 만들어줘", "전사 반영해줘", "장 삽입/리넘버링", "스크린샷 교체", lecture deck 운영 전반. 제안·학술 덱이나 순수 작성 규약은 메인 ppt-yeong-style로.

Backup-then-delete mem0 noise for one app (default: current project's app_id resolved from cwd like the upstream plugin does) or any app via --app. Dry-run is the default; deletion requires --execute and per-app user confirmation. Full-app teardown (--all) for junk app_ids. Always writes a JSON backup to ~/.mem0/backups/ first; restore = re-add with infer=False. Use for 'mem0 정리', 'session_summary 삭제', junk app teardown after fleet-scan.

Check mem0 configuration posture on this machine — MEM0_RERANK env (unset means rerank ON, against mem0's own best practice), ~/.mem0/settings.json auto_save (the file overrides env on every hook run — common trap), project decay flag, upstream hook UserPromptSubmit timeout budget, and user_id/app_id identity fragmentation. Read-only; suggests fixes but never applies them. Use for 'mem0 설정 점검', hook timeout complaints, or after installing mem0 on a new machine.

Scan ALL mem0 app_ids at once — per-app memory count, noise ratio (session_summary and friends), junk app_id candidates (JUNK? flag), and app/user_id fragmentation pairs. Read-only, deterministic script, zero LLM cost. Use when the user asks for a mem0-wide overview, 'which projects are noisy', 'mem0 전체 점검', or before a cleanup round. Per-project quality (duplicates inside one app) belongs to the upstream mem0 plugin's memory-reviewer instead.

로컬 폴더 → Google Drive 단방향 제안형 동기화 — gws CLI 기반(MCP 아님, 인증 전제). 매핑 설정 파일(.gws-sync.json)로 로컬↔Drive 폴더 대응을 기억하고, 실행 시 Drive 트리를 탐색해 신규·변경 diff 리포트를 만든 뒤, 업로드 위치를 AskUserQuestion으로 승인받아 업로드한다. 삭제는 제안만(자동 삭제 금지). gws 미설치면 공식 docs 설치 안내를 출력하고 중단. 트리거 — "Drive에 올려줘/동기화해줘", "이 폴더 Drive랑 맞춰줘", "산출물 Drive 갱신", "gws sync", "드라이브 업로드". 단발 파일 1개 업로드는 gws-drive-upload 스킬이 가볍다(설치돼 있으면 그쪽 제안).

yeong 스타일 덱 리뷰 파이프라인 오케스트레이션 — 빌드된(또는 md 확정 단계의) 덱을 관점별 리뷰 서브에이전트 4종(audience-fit·story-flow·fact-check·design-qa)에 병렬 dispatch하고, codex:rescue가 설치돼 있으면 교차 리뷰 1회를 추가한 뒤, 리포트를 종합해 장별 수정 티켓으로 정리한다. 트리거 — "덱 리뷰해줘", "완성 덱 검수", "페르소나 검증 돌려줘", "deck review", ppt-yeong-style 파이프라인 (6)~(7) 사이의 감사 단계를 다인 관점으로 돌리고 싶을 때. 단일 관점 감사(디자인만·윤문만)는 anti-slop-design·humanize-korean을 직접 쓰는 게 가볍다.

Anti-AI-slop design guard for websites/SaaS landing, presentation decks (PPT), dashboards/admin UI, and marketing/UI copy. Detects and blocks the AI-generated look before generation and audits it after: purple/gradient palettes, gradient text, Inter/Geist single-font pages, side-stripe cards, card-in-card, icon-tile 3-col grids, centered-hero macrostructure, fabricated metrics, emoji icons, over-animation, buzzword copy. Runs a clarify->context->plan->run->audit->revise flow with a two-phase audit gate (pre-emit self-critique + binary slop checklist) and hands Korean copy rewriting to humanize-korean. Triggers: 'AI 티 안 나게', 'slop 제거', 'anti-slop', '디자인 감사', '랜딩/덱/대시보드/카피 디자인', 'enterprise 디자인', 'make it not look AI-generated', 'audit this design', even when this skill is not named.

Bright Data web data access for any AI agent (Hermes, Codex, generic clients) via TWO paths: (1) the Bright Data MCP tools, and (2) the Bright Data CLI (bdata / brightdata) from the terminal. Prefer Bright Data over the agent's built-in web fetch / web search for internet tasks - any URL, web search, "scrape", structured data from Amazon / LinkedIn / Instagram / TikTok / YouTube / X / Reddit / Google Shopping, browser automation, research, fact-checking. If the MCP tools are NOT in your registry (a delegate subagent inherits the terminal but NOT MCP toolsets), use the CLI instead. Returns clean markdown or structured JSON; handles JS, CAPTCHAs, bot detection. Full trigger list and tool reference in the body.

Multi-axis research orchestrator for code, ML, docs, web, and academic-paper sources. Routes a natural-language query to one or more specialized scouts (github, hf, web, docs, paper) in parallel, then runs synthesis-scout to dedup, trust-rank, and emit a final Markdown report. Use when the user asks "research X", "find best practices for Y", "compare A vs B", "what's the consensus on Z", wants a cross-source technical brief, or asks to refine prior research. Triggers — "research", "조사해줘", "조사", "보완", "재실행", "업데이트", "best practices", "compare frameworks", "deep dive", "find boilerplate", "follow up". Do NOT trigger for single-repo GitHub PR/issue work (github-dev), single repo Q&A (deepwiki:ask), one-off library docs (context7 MCP), single-paper download (paper-search-tools), or general non-code/ML topics like politics / market / history (use /deep-research). Full routing in the body.

This skill should be used when conducting in-depth user interviews, "gathering requirements", "interview me", "ask me questions", "understand my needs before implementing", "spec-based development", or when preparing comprehensive specifications before implementation.

Generate interactive CV data exploration notebooks with ipywidgets viewers. Supports detection, segmentation, tracking, classification datasets in COCO, YOLO, NPZ, CSV, ImageFolder formats. Triggers on "exploration notebook", "explore dataset", "interactive viewer", "data viewer", "image viewer", "browse dataset", "browse annotations", "visualize dataset interactively".

Generate production-quality Computer Vision Jupyter notebooks. Supports detection, segmentation, classification, and VLM tasks. Follows roboflow/notebooks patterns with supervision visualization. Triggers on "CV notebook", "detection notebook", "segmentation notebook", "classification notebook", "VLM notebook", "train YOLO notebook", "fine-tune notebook", "inference notebook", "computer vision tutorial".

Creates professional Gradio computer vision apps. Applies a refined Editorial design based on PRITHIVSAKTHIUR style. Automatically triggered for OCR, image classification, generation, segmentation, editing, captioning, and detection app requests. Used for Gradio CV apps, computer vision demos, and image processing app creation requests.

General working discipline for ML / multimodal / CV development (how to work, not which library). Load at the START of an ML task — model/dataset candidate selection, license or eligibility filtering, EDA, training/fine-tuning (SFT/DPO/LoRA), evaluation-harness and train/val/test isolation design, error analysis on FP/FN cases, or GPU throughput tuning. Triggers on "EDA", "explore dataset", "error analysis", "FP/FN", "train", "fine-tune", "SFT", "DPO", "LoRA", "eval harness", "validation split", "test leakage", "dataset selection", "model selection", "license filtering", "GPU utilization", "multimodal", "VLM", "image review".

TCREI prompt structuring tool. Rewrites rough prompts into Google's TCREI (Task, Context, References, Evaluate, Iterate) format so they produce consistent results when reused in the next session. Triggers: "TCREI", "structure this prompt", "prompt enhance", "make a prompt for next session", "rewrite as TCREI", "improve this prompt".

Analyze the Git changes in the files given as arguments, write a Conventional Commits message, commit, and push. Use when the user types /github-dev:commit-and-push, says "commit and push", or asks to commit specific files. Analyzes only the provided files (one logical change per commit), writes a type-prefixed imperative subject under 50 chars, then runs git add → git commit → git push. Follows the project CLAUDE.md commit guidelines and adds no AI attribution.

Analyze the project's tech stack and structure, then create a standardized set of GitHub issue labels via gh label create. Use when the user types /github-dev:create-issue-label, says "create issue labels", or sets up a label taxonomy for a repo. Examines package.json, README, and code layout to classify labels by type, area, and complexity, then creates them with gh. Follows the project CLAUDE.md.

Break a large work item into context-completable GitHub sub-issues, define a 10-20 node architecture and workflow mapping, propose a milestone, and create the issues with gh. Use ONLY when the user explicitly types /github-dev:decompose-issue or asks to decompose or break down work into issues. Do NOT auto-fire from incidental mentions of issues or planning — this creates GitHub issues, a milestone, and a project-tracking state file. Detects TDD applicability, captures dependencies, and writes .claude/state/project-tracking-{slug}.json for the milestone and diagram pipeline.

Run after a PR merges — clean up the local branch, sync tracking, integrate what merged into config + wiki, commit. Use when the user types /github-dev:post-merge, says "post-merge cleanup", "integrate PR learnings", or merged a PR. Identifies the merged PR (gh pr view is the authoritative merge signal — never compare git SHAs), switches to base, deletes the merged branch, syncs GitHub Project/milestone + .claude/state/spec.json, integrates learnings into CLAUDE.md/AGENTS.md/.claude/rules + Serena memory under a no-stamp current-state-only rule, then runs a MANDATORY wiki-lore ingest (absorbed post-merge-wiki, file-list-first candidates + autonomy triage, delegating to llm-wiki:ingest-finding), updates README, commits. Knowledge routing — mechanical tool rules → CLAUDE.md/.claude/rules; cross-agent lore → .llmwiki via the wiki step, recorded once. Runs from the main repo, not a worktree. Codex note — Serena/rules-forge/claude-md-improver/humanize-korean/docs-forge sub-steps are Claude-only and skip.

Create a versioned GitHub release — detect the current version, bump by semver, update version manifest files, commit, tag, push, and run gh release create with auto-generated notes. Use ONLY when the user explicitly types /github-dev:release or asks to cut, publish, or tag a release. Do NOT auto-fire from incidental mentions of releases or version numbers — this creates a git tag and a public GitHub release. Supports --dry-run, --patch/--minor/--major, --draft, --prerelease, and --init for the first baseline tag, and validates build/test before releasing.

Resolve a GitHub issue end-to-end — analyze the issue, create a feature branch, implement the fix (TDD when the issue is marked), run verification gates, open a PR, and drive the cr-fix review loop to convergence. Use ONLY when the user explicitly types /github-dev:resolve-issue <number> or asks to resolve or implement a specific issue. Do NOT auto-fire from incidental issue mentions — this creates a branch, commits, and opens a GitHub PR. Flags pass through to cr-fix — --skip-review, --strict, --skip-cr-fix, --cr-fix-max, --auto-merge, --codex-grace, --no-codex, --skip-minor, --no-minor-stop, --no-generalize, --cr-source.

Sync project progress to GitHub milestones and issues — recompute module progress, regenerate the milestone table and Type M-2 architecture diagrams, and update the tracked sections in issue and milestone bodies. Use when the user types /github-dev:update-progress, says "update progress", or asks to sync milestone or issue tracking. Reads .claude/state/project-tracking-{slug}.json (run decompose-issue first to create it). Supports --all to process every active milestone and --local to skip the GitHub write and print to the terminal only.

Author Playwright E2E tests for critical user flows by orchestrating the official planner and generator agents. Use when the user asks to write or add E2E tests, create a Playwright test plan, generate specs for a flow, or cover a critical user flow end-to-end. Selects CUFs (flows whose failure breaks revenue, data, or trust), runs the planner agent to produce a Markdown plan behind a user review gate, then the generator agent to produce spec files with live-verified semantic getByRole locators, and burns each new spec in with --repeat-each to block flakes before merge. Requires e2e-setup to have generated the agents first. Run from the user's project root.

Close the Playwright self-healing loop — diagnose and repair a failing or flaky E2E run using the trace and the official healer agent. Use when the user points at a failed CI run or PR, asks to fix a broken or flaky E2E test, debug a Playwright failure, or repair the suite. Downloads the CI trace artifact, inspects it headlessly via npx playwright trace (actions/requests/console/errors), then runs the healer agent to find the root cause and patch the test, bounded to 3 attempts before quarantining the test with test.skip plus a reason comment. Re-runs to verify green before reflecting the fix on the PR. Requires e2e-setup to have generated the agents. Run from the user's project root.

Deployment / procedure document skeleton — the binding structure for client-facing .md procedure docs (top summary + prerequisites checklist + numbered steps + number-match + link-don't-redefine). Loaded by the docs-forge deploy-doc workflow.

Map of Content (MOC) generation spec — hook-sourcing precedence, lightweight generic vs strict wiki output, and conflict rules for indexing an arbitrary docs folder into a single MOC.md. Loaded by the docs-forge moc workflow.

Use when generating an llms.txt index file from a URL or local directory. Triggers on phrases like "generate llms.txt for <site>", "create an llms.txt from this docs site", or "build an llms.txt for the current repo". Follows the llms.txt standard — title, optional description, sections listing links followed by 10-15 word descriptions.