LLM 與 Agent

Track token usage and costs across agents for budget management

Aggregates results from multiple sub-agent executions into a unified output.

Loop automatizado de melhoria contínua que usa o Chat RAG para identificar débitos técnicos, implementa correções, reingere a base de conhecimento e valida até eliminar 100% dos débitos.

Use when implementing features or adding functionality - dispatches incremental-tdd-agent that enforces ONE test at a time, preventing batch test writing that defeats incremental design discovery. Maintains discipline through entire feature implementation (10+ test cycles)

Store APPLICATION patterns (architecture, procedures, conventions) in AgentDB's skills table. NOT for swarm/transient memory.

Create and structure Claude Code agent skills. Use when the user wants to create a new skill, write an agent skill, make a claude skill, or asks about SKILL.md files.

Quickly scaffolds new Microsoft 365 Copilot declarative agent projects using ATK CLI. Collects basic project information and creates the initial project structure. Use only when creating a new empty M365 Copilot agent project from scratch.

根据官方规范和最佳实践评估 Agent Skill 设计质量。用于审查、审计或改进 SKILL.md 文件和 skill 包。提供多维度评分和可操作的改进建议。

Follow the project's Python-based tech stack using uv, FastMCP, Claude Agent SDK, Pydantic v2, SQLite with SQLModel, and pytest with ruff. Use this skill when writing Python code for the MCP Task Aggregator. When using FastMCP to create MCP servers or tools. When working with Claude Agent SDK for AI agents. When defining Pydantic models or using pydantic-settings for configuration. When writing SQLModel queries or database operations. When using loguru for logging. When integrating with Jira (jira-python), GitHub (pygithub), or Linear (linear-sdk). When running pytest tests or ruff linting. When configuring GitHub Actions CI/CD.

🚀 AUTONOMOUS INTELLIGENT AGENT that orchestrates the entire skill system. Proactively monitors project health, decomposes complex tasks into skill workflows, learns from execution patterns, and evolves continuously. Use when "orchestrate skills", "autonomous recommendations", "task decomposition", "system health check", "skill evolution", or analyzing skill architecture. The meta-brain of the skill ecosystem.

Production hybrid search with PGVector 0.7+ and BM25 using Reciprocal Rank Fusion, metadata filtering, and HNSW indexing for semantic retrieval

MUST be loaded when setting up, installing, migrating, reviewing, auditing, or checking CLAUDE.md files in projects. Covers installing the promode CLAUDE.md into new projects, migrating existing CLAUDE.md content to READMEs (progressive disclosure), and auditing projects for conformance. Invoke PROACTIVELY when user mentions CLAUDE.md, project setup, agent configuration, or code meta files.

Generate and validate tech stack documentation. Use when starting a new project, documenting technology choices, or when user asks to create/update tech stack documentation. Includes structural validation and LLM-as-Judge content review.

Expert implementation of @vis.gl/react-google-maps library for Google Maps in React. Use when building maps, markers, pins, infowindows, places autocomplete, geocoding, draggable markers, polygons, circles, polylines, drawing tools, or any Google Maps JavaScript API integration in React/Next.js applications.

Ensures CLAUDE.md, README.md, and CHANGELOG.md are updated before git push operations

Configure WaveCap LLM-based transcription correction. Use when the user wants to enable/disable LLM correction, change models, tune prompts, or optimize correction quality on Apple Silicon.

JavaScript 代码检查和修复工具。当需要执行 JavaScript 代码时，使用此技能进行：1) 检查本地 Node.js 版本是否支持代码语法，2) 自动修复全角符号等常见问题，3) 验证代码可执行性，4) 返回检查结果供主 Agent 使用 execute_shell_command 执行

Automatically applies when evaluating LLM performance. Ensures proper eval datasets, metrics computation, A/B testing, LLM-as-judge patterns, and experiment tracking.

Manages and synchronizes AI instructions (global and project-level) across various tools (Gemini, Claude, Copilot, etc.). Use this skill when you need to: (1) Sync `templates/global-instructions/master.md` to all tool headers, (2) Generate or update `PROJECT_INSTRUCTIONS.md` in a directory, (3) Update sub-agent instructions, or (4) Maintain consistency in AI behavioral guidelines.

Spawn a Claude worker in an isolated git worktree for parallel development. Use when you need to work on a task in isolation without affecting current work.