AI Engineering

Looper scaffolds an agent loop with goal refinement, verification criteria, reviewer and judge roles, and stop guards. Use it when designing or resuming a /goal-style workflow or cross-model review loop.

Agent Teams coordinates a default roster of Claude Code agents for feature work. It enforces a fixed TDD pipeline from spec and tests through review, security checks, and PR creation.

Agentic Development builds autonomous AI agents that plan, use tools, and verify multi-step tasks. It uses Pydantic AI for Python and Claude Agent SDK for Node.js/Next.js projects.

AI Models Reference lists current Claude, OpenAI, Gemini, Eleven Labs, Replicate, and related model choices with example APIs. Use it when you need to pick a model for reasoning, code, vision, embeddings, voice, or image generation.

Council Review validates plans, architecture changes, and PRs with a multi-model panel before execution. Use it when you want independent review through validate-plan or review workflows.

Database Schema Awareness reads the schema before database work so you verify columns, types, and relationships before writing queries or models. Use it to prevent column-name and type mismatches and to guide generated types.

External Model Delegation routes each user prompt to a chosen model or CLI based on cost and task type. Use it to delegate simple, medium, or complex requests to local or external models from a Claude Code hook.

Run History Skill Builder turns completed tasks, browser flows, logs, or recovered workflows into a new reusable skill package or a reviewed design plan. Use it when you need to extract a workflow from evidence, summarize lessons, or hand off an upgrade for an existing skill.

Run History Skill Upgrader plans structural improvements for an existing skill using run evidence, validation failures, source drift, platform drift, and user feedback. It is used to propose and, after explicit approval, apply changes when a skill needs maintenance or cleanup.

Session Handoff Prompt creates a paste-ready continuation prompt for resuming a long or degraded agent session. Use it when starting a fresh session needs the current objective, verified facts, and next actions preserved.

Task Clarifier asks structured questions to resolve a user's goals, constraints, and success criteria before work begins. Use it when a request is vague, conflicting, or needs explicit confirmation.

User Profile Keeper maintains a local persistent user profile for agent harnesses. Use it to create, update, query, correct, delete, export, or audit profile facts and collaboration preferences when explicitly invoked.