Senior code reviewer for Python projects with emphasis on agentic/LLM systems. Reviews for correctness, design clarity, maintainability, testability, observability, and idiomatic LangGraph usage. Produces a structured review report. Read-only; does not modify code.
Documentation specialist for README, ADRs, architecture diagrams (mermaid), assumptions logs, and "Additional Capabilities" sections. Use when producing or revising any markdown artifact intended for maintainers, collaborators, or users. Writes only to markdown files; every edit is ask-gated.
Senior LangGraph architect for state-schema design, StateGraph topology, conditional routing, interrupt/resume semantics, checkpointer choice, reducer selection, structured outputs, and tool/MCP wiring. Use for designing or reviewing a graph before implementation, or auditing a graph for correctness. Read-only.
Senior security engineer specialized in vulnerability detection, threat modeling, secure coding, and ML/LLM/agent security. Covers web, API, infrastructure, supply chain, model integrity, prompt injection (direct and indirect via RAG/tools), agentic tool-calling risks, and MCP server trust. Use for security-focused…
Senior QA and ML/LLM Test Engineer specialized in test strategy, test writing, coverage analysis, regression tests, data and feature contracts, ML pipeline validation, and LLM/agent/RAG evaluation. Use for designing test suites, writing tests for existing code, building bug-reproduction tests, evaluating coverage, or…
Canonical LangGraph patterns for typed State, conditional routers, interrupt/resume, checkpointers, structured outputs, tool/MCP wiring, subgraphs, parallel branches, async streaming, and LangSmith tracing. Use when designing or implementing any LangGraph node, router, or graph. Reference; not a runbook.
Standard Python project setup using uv or conda + pip, src-layout, ruff (lint + format), mypy strict, pytest + pytest-asyncio, and pre-commit. Includes pyproject.toml, uv, and environment.yml templates, a GitHub Actions CI snippet, and LLM-friendly test patterns. Use when scaffolding a new project or upgrading…