Ai plugins

Context management and multi-agent orchestration

2 commands · 1 agent

AI & Agents bundle — 388 curated skills (agents, ai-ml, prompt-engineering) for Claude Code, Codex & Antigravity.

200 skills

Gemini CLI integration for Claude Code — use Google's Gemini models from within Claude Code

3 skills · 7 commands · 1 agent · hooks

Claude Code Plugin for Nano Banana (Gemini 2.5 Flash Image) model - generate and manipulate images with text prompts

8 commands · 1 MCP server · hooks

Subagents on ChatGPT/Codex GPT-5.6 models (Luna, Sol, Terra), routed through shunt at the inference layer.

5 agents

Wire AnyRouter, an OpenAI-compatible LLM gateway, into projects and agents, with migration help and an MCP server.

1 skill · 1 command · 1 agent · 1 MCP server

Route Claude Code to 19 LLM backends (local Ollama/LM Studio, cloud OpenAI/Anthropic/Gemini/Groq/DeepSeek/Mistral/etc., OCI Generative AI, and CLI agents Claude Code/Codex [gpt-5-5]/Gemini/Antigravity [Gemini 3.x]) through one local Anthropic-compatible gateway. Beats-frontier model fusion via the `fusion` / `auto` slugs (panel + judge synthesis with a generous token budget for the best answer), log-driven smart routing, cost prediction and budgets with alerts, quota-aware failover, council/2nd-opinion, shared cross-LLM memory, a result cache, and a real-time dashboard with cost/forecast/budget/routing panels. Self-bootstrapping (installs an isolated venv on first run). Tenancy-agnostic and bring-your-own-env: no keys, OCIDs, endpoints, or tenant data shipped — each user defines OCI profile, API keys, and (optional) Langfuse/OCI APM telemetry in their own environment.

Write documents for agents, analyze a skill, prune the prose that does not change behavior, split broad references into focused files, and keep skills lean as they change

5 skills

AI and Data Engineering: data pipelines and warehousing, AI/ML landscape and production patterns, RAG and AI agent development, and LLM token and cost optimization on Azure (including provider-agnostic prompt/context compression).

4 skills

AI Chatbot Strategy: foundational AI and chatbot concepts for business, RAG architecture explained for non-technical decision-makers, custom chatbot build and deployment guide, and chatbot business value with ROI frameworks.

4 skills

Machine Learning: problem framing and when not to use ML, data work and leakage, the classical toolkit (gradient boosting, linear models, clustering), deep learning fundamentals and architectures (transformers, CNNs, diffusion), the PyTorch/JAX ecosystem and torch.compile, training at scale (DDP, FSDP, tensor/pipeline parallelism, mixed precision), fine-tuning and PEFT/LoRA, honest evaluation, debugging training, inference and serving (vLLM, SGLang, quantization, KV cache), MLOps and reproducibility, hardware and cost, and interpretability, fairness, and safety.

4 skills

Retrospect on a coding session to improve the environment the next one runs in

1 skill

Hand a side task found in an agent's thread to a new agent in its own Herdr worktree, with a brief that gives it the thread's context

1 skill

State-of-the-art prompt and context engineering toolkit for 2026. Includes DSPy-inspired optimization, Claude 4.x patterns, context engineering, chain-of-thought design, RAG prompting, multi-agent orchestration, and model-specific guidelines for GPT, DeepSeek, Qwen, Grok, Kimi K2, and more.

18 skills · 3 agents

A repo-native toolchain for LLM apps: scaffolds eval suites, red-teams for prompt injection, hardens structured-output reliability, and audits token cost, riding promptfoo/DeepEval/Pydantic rather than replacing them

4 skills · 4 commands · 2 agents · hooks

Generate and edit AI media (image, video, audio, 3D) on WaveSpeed through its hosted MCP server and a CLI skill for local files, batches, live catalog search and price quotes.

1 skill · 1 MCP server

AI 图像提示词生成技能。基于 Nano Banana Pro 的 20 条核心提示技巧，输出 YAML 格式的分层提示词结构

1 skill

Self-evolving research methodology: 18 universal principles for LLM-driven investigation, distilled from a 376-turn session with 1 positive + 17 null campaigns. Covers epistemic foundations (causality, nulls, record-keeping), investigation patterns (multi-lens agents, adversarial gates, per-trade enrichment), decision layer (constraint-pivot, supersede-not-rewrite), and emergent resurrection (failure archive with resurrect_if conditions).

5 skills

AI prompt generation and review tools for LLM development

2 commands

Gemini Deep Research via browser automation: Playwright CDP, 8-step flow, 40k+ char research reports

1 skill

Find voices, generate speech, transcribe audio and benchmark AI models with Inworld.

24 skills · 1 MCP server

Build AI systems. Agents: loop-design-check, agent-architecture-audit, autoresearch-agent, cli-for-agents, writing-for-agents and llm-friendly-context, model-router. Skills and servers: skill-creator, mcp-builder, harvest and book-to-skill. Fleets: herdr-orchestrator, herdr-worker, herdr-atomic and herdr-setup run a multi-agent worker fleet through the horch CLI. Models: decide whether to fine-tune, build the eval harness first, curate datasets, LoRA/QLoRA, preference optimisation, GRPO/RLVR, checkpoint promotion, and RAG and recommender pipeline design. skill-creator and mcp-builder are owned copies of Apache-2.0 skills from anthropics/skills; vendored skills are credited in NOTICE.md.

24 skills

Thin Claude Code skills over the Kagura Memory CLI (kagura): doctor, auth, setup, memory (remember/recall), document ingestion, resource tokens, R2 file uploads, and zero-knowledge secrets.

9 skills

Route each task to the cheapest capable model, grade what comes back, and recalibrate the matrix from those grades.

1 skill · 3 commands · 1 MCP server

Query any opencode-supported model for a second opinion, or delegate a task on a public repo to a private headless opencode instance — all from inside Claude.

1 skill · 1 agent · 1 MCP server

Query GitHub Copilot models programmatically via the official SDK — no token setup required, uses your logged-in Copilot CLI credentials.

1 skill · 1 MCP server

Generate and edit images with GPT, FLUX, and MAI in Microsoft Foundry, with optional resumable comparisons and offline delivery.

1 skill

Evolver for Claude Code, from a hooks module with no tools: recalls one EvoMap strategy per prompt, reports each reuse with the evidence the turn produced, and distills verified work into drafts you publish or discard.

Expert skills, agents, and commands for LlamaCloud (Parse v2, Extract v2, Classify, Split, Sheets, Cloud Index), LiteParse, the LlamaIndex framework, and agentic Workflows. Wires the official LlamaIndex docs MCP for live fact lookups.

13 skills · 2 commands · 5 agents · 1 MCP server

LLM application, agent, MCP server, evaluation, and model training skills

20 skills