LLM & Agents

Cross-AI CLI communication for Kitty terminal. Use when collaborating with other AI assistants (Codex, Gemini, Aider, Cursor, OpenCode), delegating tasks, or requesting code review.

Invoke for complex multi-step tasks requiring intelligent planning and multi-agent coordination. Use when tasks need decomposition, dependency mapping, parallel/sequential/swarm/iterative execution strategies, or coordination of multiple specialized agents with quality gates and dynamic optimization.

Python unit testing patterns for pytest, including test file structure, fixtures, mocks, and coverage strategy. Use when asked to write tests (unit/integration) for any Python module, function, class, or method.

This skill should be used when the user asks about "LLM providers", "model configuration", "streaming", "prompts", "token optimization", "AI SDK", or needs to work with language model integration in XSky.

Claude Code mastery for power users. Build agents, MCP servers, publish plugins.Use when: user wants to create agents, build integrations, or publish plugins.

Routes queries to appropriate data sources (vectordb, web_search, direct_llm) and selects collections to search. Use when analyzing user queries, determining search strategy, or when the user mentions routing, collections, or data source selection.

This skill provides knowledge for generating effective Claude Code agents tailored to specific projects. It is used internally by the agent-team-creator plugin when analyzing codebases and creating specialized agent teams. Contains templates, best practices, and patterns for writing project-aware agents.

Execute implementation plan by dispatching fresh subagent for each task, with code review between tasks

pgvector - PostgreSQL extension for vector similarity search. Use for embedding storage, cosine similarity, IVFFlat indexes, and HNSW indexes.

Comprehensive expertise for working with Microsoft's GenAIScript framework - a JavaScript/TypeScript-based system for building automatable LLM prompts and AI workflows. Use when creating, debugging, or optimizing GenAIScript scripts, implementing prompts-as-code, working with tools and agents, processing files (PDF, CSV, DOCX), defining schemas, or building AI automation workflows.

전체 SEMO 패키지 통합 품질 감사. Use when (1) 전체 패키지 품질 점검, (2) Agent/Skill/Command 표준 준수 검토, (3) 비효율적 구조 탐지, (4) 정기 감사 수행.

OPA resource governance policies enforcing CPU/memory limits, ResourceQuota compliance, LimitRange validation, and storage constraints.

Systematically diagnoses and fixes failures through hypothesis-driven investigation. Auto-activates when user reports something broken, failing, erroring, not working, crashed, or buggy. Spawns debugger agent for binary search isolation and minimal reproduction.

Use when creating or editing skills, before deployment, to verify they work under pressure and resist rationalization - applies RED-GREEN-REFACTOR cycle to process documentation by running baseline without skill, writing to address failures, iterating to close loopholes

Use when creating or developing, before writing code or implementation plans - refines rough ideas into fully-formed designs through collaborative questioning, alternative exploration, and incremental validation. Don't use during clear 'mechanical' processes

Spawn threads on running claude-threads orchestrator

Creates Claude skills with proper structure and best practices. Use when the user asks to create a new skill, generate a skill, or build a custom Claude capability.

Reduce LLM API costs without sacrificing quality. Covers prompt caching (Anthropic), local response caching, prompt compression, debouncing triggers, and cost analysis. Use when building LLM-powered features, analyzing API costs, optimizing prompts, or implementing caching strategies.

This skill provides knowledge about Claude tools (Prompts, Skills, Subagents, Commands, and MCPs). Use this skill when knowledge about Claude tools is needed, when the user mentions Claude tools or when the user needs help identifying the best Claude tool to use/create for a task. Common questions include 'Should I create a Claude tool for this use-case?', 'What type of Claude tool should I create for this use-case?."

Run repo startup: prompt load, checklist sync, optional PR triage, worktree hygiene, and CI pull. Best-effort with structured status.