tool-use plugins

by Matt Corbett

AI agent-engineering team — agents (agentic-systems-architect, agent-implementation-engineer) for BUILDING production agentic systems: hard TRIAGE (most tasks win as a fixed workflow or a single LLM call, not an agent), single-vs-multi-agent topology, framework selection (LangGraph / OpenAI Agents SDK / CrewAI / AutoGen / Claude Agent SDK), tool & function design, context/memory/state management, planning & reflection loops, human-in-the-loop, guardrails, cost/latency budgets, tracing/observability, and agent EVALS (trajectory, tool-use, task-completion). skills, a knowledge bank, and templates. Engineering judgment, not a demo; the framework/model/API landscape is volatile — every version carries a retrieval date + [verify-at-use]. Distinct from ai-rag-engineering (retrieval), prompt-engineering (single-call prompt craft), llm-evaluation-engineering (general eval harnesses), and claude-app-engineering (Claude product apps). Needs ravenclaude-core.

by Matt Corbett

Claude app-engineering specialist team — agents (claude-solution-architect, prompt-and-context-engineer, mcp-and-server-tools-engineer, agent-sdk-engineer, eval-engineer, claude-app-ops-engineer) for building production applications on the Claude API, the Claude Agent SDK, and MCP. Covers the build-surface decision (Messages API / Agent SDK / Managed Agents), model right-sizing (Opus 4.8 / Sonnet 4.6 / Haiku 4.5, 1M context), prompt caching, tool use + structured output, MCP server authoring + Anthropic-hosted server tools (computer use, code execution, Files API, memory), evals (golden sets, LLM-as-judge), and FinOps + reliability (batch, model routing, backoff, observability). dated-and-cited knowledge bank, templates, an advisory hook (house opinions). Seams: prompt-as-artifact, security, and cross-domain architecture escalate to ravenclaude-core. Requires ravenclaude-core@>=0.7.0.

by Matt Corbett

Prompt & context engineering team — agents (prompt-architect, prompt-implementation-engineer, prompt-reliability-engineer) that own the layer between the model and the app: prompt & context strategy (task decomposition, pattern selection across few-shot / CoT / decomposition / role, system-vs-user structure), the output-format contract (JSON mode, tool/function-calling schemas, few-shot curation, token budgeting), and prompt reliability (eval sets & regression, prompt versioning/CI, prompt-injection & jailbreak defense at the prompt layer). Provider-neutral. skills, a knowledge bank with Mermaid decision trees + a dated 2026 reference, best-practices, templates. Seams: which model → ai-coding-model-guidance / claude-api; retrieval → ai-rag-engineering; eval at scale → llm-evaluation-engineering; adversarial testing → ai-red-teaming; the app → claude-app-engineering / backend-engineering. Requires ravenclaude-core@>=0.7.0.