azadehtavassoli
10 verified skills
topics/agent
# AGENT TOPIC SKILL ## Purpose Define agent systems as deterministic execution engines with explicit control, observability, and schema-validated interfaces. ## When to Use - Building single-agent or multi-agent systems - Designing tool-enabled execution workflows - Implementing framework adapters for agent runtimes ## Hard MUST Rules 1. Agent execution MUST be deterministic under fixed inputs/config. 2. LLM is an optional component; business logic MUST remain executable with stubs. 3. Tool r
tools
topics/fastapi
# FASTAPI TOPIC SKILL ## Purpose Provide deterministic backend API patterns with typed schemas, layered architecture, and auditable behavior. ## Hard MUST Rules - Pydantic schemas required for request/response. - Error handling and logging must be explicit. - Settings/environment handling must be typed and validated. - Tests must be deterministic and isolated.
development
topics/mcp
# MCP TOPIC SKILL ## Purpose Implement Model Context Protocol (MCP) servers and clients with the official MCP Python SDK (`mcp`) using deterministic, transport-explicit patterns. ## When to Use - Building MCP servers over `stdio` for local process integration - Building MCP servers over Streamable HTTP for remote/networked access - Building MCP clients that connect to local (`stdio`) or online (`streamable-http`) servers - Defining reusable MCP templates and transport-specific implementation p
tools
topics/document_intelligence
# Document Intelligence Skill ## Purpose Convert unstructured documents into LLM-ready Markdown using MarkItDown, with deterministic defaults and governance-compliant wrappers. References: - `core/GLOBAL_RULES.md` - `core/LOGGING_STANDARD.md` - `core/TESTING_STANDARD.md` - `core/STRUCTURED_OUTPUT_STANDARD.md` ## When to use - You need a standard document-to-Markdown conversion path. - You need deterministic conversion that does not require an LLM by default. - You need optional visual/LLM int
development
topics/ui
# UI TOPIC SKILL ## Purpose Define deterministic UI implementation patterns with accessibility, component decomposition, and reproducible behavior. ## Framework Adapters - `topics/ui/frameworks/streamlit/` - `topics/ui/frameworks/gradio/` ## Hard MUST Rules - Component contracts must be explicit. - Accessibility checks are mandatory. - UI state transitions must be predictable and testable. - Framework-specific UI logic MUST remain inside adapter folders.
development
topics/a2a_server
# A2A SERVER TOPIC SKILL ## Purpose Implement Agent2Agent (A2A) servers with the official Google-maintained `a2a-sdk` using explicit request handling, task state, and deterministic execution contracts. ## When to Use - Creating a new A2A-compatible server - Exposing agent capabilities through an A2A Agent Card - Implementing executor/request handler patterns for A2A workflows ## Hard MUST Rules 1. Use official `a2a-sdk` server modules for handlers, tasks, and app wiring. 2. Keep executor logi
development
topics/langraph
# LangGraph Skill ## Capability You can design and build stateful, long-running agents using LangGraph's low-level orchestration primitives and integrate observability via LangSmith for development and debugging. ## Key Abilities ### 1. Design Agents with LangGraph - **Graph-based state machines**: Design agents as directed acyclic graphs (DAGs) using nodes (computation steps) and edges (transitions between them). - **Explicit control flow**: Define precisely what happens at each step—no hid
tools
topics/rag
# RAG TOPIC SKILL ## Purpose Define deterministic retrieval-augmented generation pipelines with grounded outputs and measurable evaluation. ## Hard MUST Rules 1. Retrieval, ranking, and generation stages MUST be explicit and logged. 2. Groundedness and citation integrity MUST be validated in tests. 3. Evaluation outputs MUST be structured and reproducible. 4. No real LLM/provider calls in automated tests. 5. Structured output models are mandatory.
testing
topics/demo_pipeline
# Demo Pipeline — SKILL ## What this skill is for Use this skill to implement a long-term, maintainable demo pipeline for product demos. The aim is to create a reusable machine for demos, not a one-off screen recording. ## Canonical design A good demo pipeline has these responsibilities: ### 1. Scenario definition A scenario defines: - objective - audience - core message - preconditions - interaction steps - hero moment - output profiles Use: - `brief.md` for human-readable story and intent
development
topics/a2a_client
# A2A CLIENT TOPIC SKILL ## Purpose Implement Agent2Agent (A2A) clients with the official Google-maintained `a2a-sdk`, using modern client factory APIs and deterministic message handling. ## When to Use - Building clients that connect to A2A-compatible agents - Sending task/message requests to local or remote A2A servers - Creating reusable A2A client adapters in Python ## Hard MUST Rules 1. Use official `a2a-sdk` client APIs (`ClientFactory`, `ClientConfig`, helper builders). 2. Prefer moder
tools