Optimize AgentDB performance with quantization (4-32x memory reduction), HNSW indexing (150x faster search), caching, and batch operations. Use when optimizing memory usage, improving search speed, or scaling to millions of vectors.
Review an existing Python RAG codebase against an explicit RAG architecture document and produce a production-readiness backlog with priorities, rationale, and execution guidance.
Specialized skill for ML training workflows on cloud GPUs. Fine-tune LLMs with LoRA/QLoRA, train image LoRAs, build classifiers, and run custom training jobs. Generates production-ready training pipelines with checkpointing, logging, and optimal GPU selection.
Create and configure custom subagents in Claude Code. Use when user asks to create a new subagent, needs specialized AI assistance with specific tool restrictions or permissions, implements task-specific workflows with isolated context, sets up domain-specific agents with custom prompts, or creates subagents with hooks or validation rules
Use this skill when the user needs help writing, understanding, or modifying ROS launch files. Covers launch XML syntax, arguments, parameters, node configuration, includes, conditionals, namespaces, remapping, and common patterns. Trigger examples: - "How do I create a launch file?" - "Add an argument to this launch file" - "What does $(find pkg) do?" - "How do I include another launch file?" - "Set up conditional launching" - "Configure node parameters in launch"
Multi-LLM workflow patterns for complex reasoning tasks. Based on Anthropic's "Building Effective Agents" research. Use this skill when: (1) Tasks require multiple perspectives or approaches (2) Complex decisions need structured analysis (3) Problems benefit from decomposition into subtasks (4) Quality requires iterative refinement (5) Input needs routing to specialized handling Patterns: chain, parallel, route, orchestrator-workers, evaluator-optimizer
Capture decisions and requirements from conversation history. Use when user says "harvest", "capture decisions", or "save what we discussed" to persist architectural decisions into .rivet/systems.yaml.
Link triplet entities to soul nodes - connects knowledge layers
Best practices for crafting effective prompts for agents, skills, and Claude interactions. Use when writing system prompts, agent instructions, or optimizing Claude responses.
Guide for writing effective prompt specifications (specs) for the GPT-5.2 Codex agent. Use when the user asks to "write a prompt spec", "create a system prompt for GPT-5.2", "design a spec for an agent", or "how to prompt GPT-5.2".
Generate a complete OAF-compliant agent package from an Agent Case (requirements) and Agent Design (architecture). Use when asked to "package an agent", "generate an agent package", "create an agent from a design", or when you have both case/ and design/ folders and need to create the package/ folder with AGENTS.md, skills, and scripts.
Universal Reasoning Framework implementing λο.τ calculus over holarchic structures. Provides severity-based routing (R0-R3 pipelines), modular cognitive architecture (DEC, EVL, PAT, SYN, MEA, HYP, INT), fractal execution patterns, multi-level validation (η≥4, KROG), and adaptive learning. Triggers on: (1) complex multi-step reasoning, (2) high-stakes decisions requiring validation, (3) research synthesis across domains, (4) system design and architecture, (5) crisis management, (6) performance optimization. Implements scale-invariant reasoning from micro (tool calls) through meso (skill composition) to macro (orchestrated workflows).
Generate or edit images using Gemini 3 Pro Image via OpenRouter API. Use when the user wants to: (1) Generate images from text prompts, (2) Edit/modify existing images, (3) Compose/blend multiple images. Triggers: "generate image", "create picture", "edit photo", "make an image".
ScreenApp multimodal video/audio analysis CLI with graph-based context retrieval
Gemini任务委派专家。当需要前端UI/UX、样式设计、React/Vue组件、CSS优化时自动激活。帮助构建正确的Gemini调用。
Multi-agent deliberation protocol with persona-based debate, evidence-weighted consensus, and adaptive stopping. Use for complex reasoning requiring diverse perspectives, stress-testing decisions, or when single-viewpoint analysis is insufficient. NOT for simple lookups, routine tasks, or time-critical responses.
This skill defines the Initializer Agent protocol for the Two-Agent Pattern. The Initializer Agent runs ONCE at the start of a new task to analyze, plan, and set up state for Coding Agents. Use this when starting a new autonomous development task that requires session boundaries.
This skill should be used when users need to rent GPUs, run ML training jobs, or manage compute resources on Basilica's decentralized GPU marketplace. Use it for PyTorch/TensorFlow training, distributed training setup, GPU rental management, cost monitoring, or any Basilica CLI workflows. Includes workaround for non-TTY environments like Claude Code.
Quality-gated iteration with multi-agent workflows for complex tasks
Invokes 300+ AI models via OpenRouter API for text completion, image generation, and model discovery. Use when delegating tasks to external models (GPT-5.2, Gemini 3, Llama, Mistral, etc.). Triggers on "use OpenRouter to...", "call GPT-5 to...", "generate an image with Gemini", or similar requests for external AI models.
Comprehensive guide to securing LLM applications including prompt injection prevention, jailbreak detection, guardrails, and red teaming methodologies
最高品質のSkillを設計・生成するメタスキル。新しいSkillが欲しい、繰り返し作業を自動化したい、ベストプラクティスに準拠したSkillを作りたいときに使う。
