AI Engineering

CLI Chat sends chat completions, streams responses, and starts an interactive REPL session from the command line. Use it to work with OmniRoute providers, combo routing, and system prompt configuration.

OmniRoute A2A CLI sends tasks to the OmniRoute A2A server and inspects skill execution history. Use it to test the JSON-RPC 2.0 agent-to-agent protocol interactively from the command line.

Creates slide decks from text, outlines, documents, or a reference image using the 2slides API. It can also add voice narration and export slides as images or audio.

Routes a goal through acc's scored-memory loop, deliberates on returned brain_frames, and submits proposals to continue the loop. Use for agentic problem-solving with memory scoring.

Implements LLM-as-judge techniques for evaluating model outputs with direct scoring, pairwise comparison, and bias mitigation. Use when building automated evaluation pipelines or comparing AI-generated responses.

Provides expert legal analysis and strategy for cases involving Maria da Penha Law, domestic violence, feminicide, and Brazilian criminal procedure. Useful for victims, defendants, lawyers, and students seeking technical legal guidance.

Designs, writes, and deploys new Claude Code skills following the Diamond Standard. Use this skill to generate custom, expert-level agent capabilities with a structured workflow.

Diagnoses AI project workspaces and creates optimal technical roadmaps, either from scratch or for existing codebases. Best used at project inception to plan tasks and assemble an expert AI squad, with all communication in Spanish.

Triage open promises and close them with honest, real-world verdicts using acc_act(runtime='outcome'). Ideal for agentic workflows tracking commitments.

Analyzes a project's niche to inject domain-specific knowledge, regulations, and standards. Use after identifying the niche to align product, design, and technology with industry requirements.

Orca Orchestration coordinates structured multi-agent work through threaded messages, blocking ask/reply flows, task dispatch, waits, task DAGs, decision gates, and coordinator loops. Use it when you need to supervise, monitor, wait for results, or decompose work across agents rather than hand off full ownership.

Academic Paper Reviewer simulates a full international journal peer review process with five dynamically configured reviewer agents, producing an editorial decision and revision roadmap. Use it before submitting a manuscript or to check whether revisions address reviewers' comments.