发现 AI Agent Skills

Weglot integration. Manage data, records, and automate workflows. Use when the user wants to interact with Weglot data.

Docker and Docker Compose patterns for local development, hardened CLI installer harnesses, container security, networking, volumes, and multi-service orchestration. Use when creating or reviewing Dockerfiles and Compose services, testing installers across Linux distributions, or planning accurate native macOS and Windows validation.

Train and fine-tune transformer language models using TRL (Transformers Reinforcement Learning). Supports SFT, DPO, GRPO, KTO, RLOO and Reward Model training via CLI commands.

Use when a coding session is burning context fast, output is drowning in noise, the model keeps forgetting earlier decisions, or the user asks to reduce tokens, stop the verbose narration, work in a huge repo, or make a long session survive. Covers what to read, what to summarize, what to write to disk, and when to start fresh.

Validates that a DAG node's output matches its declared JSON schema before passing to downstream nodes. The glue that makes multi-agent DAGs reliable. Use when checking output contract compliance, generating validation schemas from descriptions, or debugging contract mismatches between nodes. Activate on "validate output", "output contract", "schema validation", "contract mismatch", "output doesn't match". NOT for content quality assessment (use dag-quality), skill grading (use skill-grader), or general JSON schema work outside DAG context.

睡前/离开前额度耗尽时，设置一次性本机定时器，在 Claude 订阅额度重置后自动发一条极简 claude -p 消息，立刻开启新的 5 小时用量窗口——用户睡醒时新窗口已跑掉大半，等于白赚一个 冷却期。Use when 用户说「还有 X 小时重置额度，帮我定时戳一下 Claude」「我去睡觉了， 到点 ping 一下开启新窗口」「额度用完了，重置后帮我踩个点」「设个定时器开启新的 5 小时 冷却期」，或提到 定时戳、戳 Claude、quota ping、开窗、额度重置后自动发消息。 仅限一次性定时任务与 macOS 本机（caffeinate/BSD date）；周期性任务用 /loop 或 cron， 查询 ChatGPT/Codex 何时重置用 tibo-reset-codex，分析额度为什么烧完用 claude-usage-analyst。

MANDATORY recipe for every Caffeine build that reads Spotify catalog data or drives a user's Spotify account from a canister. The supported path is the `spotify-client` mops package (Spotify Web API) over outbound HTTPS with an OAuth 2.0 bearer token minted off-chain. Hand-rolling `ic.http_request` calls to `api.spotify.com` is a FORBIDDEN anti-pattern — it bypasses the non-replicated-outcall safeguard (player state and `progress_ms` differ per node and fail consensus), the generated JSON decoding, and the bearer handling. Load this skill whenever the user, spec, or any prior task mentions Spotify, a song, track, artist, album, playlist, music search, new releases, genres, markets, "what's playing", recently played, the queue, player controls (play / pause / skip / shuffle / repeat), a saved library, podcasts (shows or episodes), chapters or audiobooks — and BEFORE writing any code that touches a Spotify endpoint.

Create, generate, or scaffold Makefiles with .PHONY targets and build automation.

Generate audio visualization videos using each::sense AI. Create waveforms, spectrum analyzers, particle effects, 3D visualizations, and beat-synced animations from audio files.

Google Calendar: Create a new event.

When the user wants to write content for a LinkedIn carousel, Instagram carousel, Facebook carousel, TikTok photo carousel, Pinterest Idea Pin, or any swipeable multi-slide format. Also use when the user mentions 'carousel,' 'slides,' 'LinkedIn carousel,' 'Instagram carousel,' 'IG carousel,' 'photo carousel,' 'TikTok photo carousel,' 'Idea Pin,' 'Pinterest Idea Pin,' 'swipe post,' 'slide deck,' or 'visual content.' Outputs slide-by-slide text content (not visual design). For single posts, see post-writer-sms. For threads, see thread-writer-sms. For caption copy under each slide post, see caption-writer-sms.

APImetrics integration. Manage data, records, and automate workflows. Use when the user wants to interact with APImetrics data.

Master prompt engineering for AI models: LLMs, image generators, video models. Techniques: chain-of-thought, few-shot, system prompts, negative prompts. Models: Claude, GPT-4, Gemini, FLUX, Veo, Stable Diffusion prompting. Use for: better AI outputs, consistent results, complex tasks, optimization. Triggers: prompt engineering, how to prompt, better prompts, prompt tips, prompting guide, llm prompting, image prompt, ai prompting, prompt optimization, prompt template, prompt structure, effective prompts, prompt techniques

Set up, adopt, refresh, or audit a repository's agent instructions (the AGENTS.md block) so AI agents work well in that repo. Also records observed agent mistakes as pitfalls. Use when invoked by name

Configure use when configuring load balancers including ALB, NLB, Nginx, and HAProxy. Trigger with phrases like "configure load balancer", "create ALB", "setup nginx load balancing", or "haproxy configuration". Generates production-ready configurations with health checks, SSL termination, sticky sessions, and traffic distribution rules.

Use when choosing how to integrate Supabase into a specific stack — setting up Next.js SSR auth flows, wiring an SPA or React Native client, configuring mobile deep links, or designing multi-tenant data isolation. Covers where the client runs (browser vs server) and which key it uses (anon respects RLS, service_role bypasses it). Trigger with phrases like "supabase next.js", "supabase SSR", "supabase react native", "supabase SPA", "supabase serverless", "supabase multi-tenant", "supabase server component", "supabase architecture", "supabase service_role server".

Add a restrained particle atmosphere inside one section with configurable shapes, density, gravity, wind, sway, rotation, recycling or settling, pointer disturbance, visibility pausing, responsive limits, and reduced-motion fallbacks. Use for petals, leaves, snow, sparks, confetti, dots, paper, icons, or brand fragments that support a section's mood without obscuring content.

[Utilities] Use when you need claudeKit usage guide - just type naturally.