Skill UI

A quota-aware parallel subagent coordination skill for Antigravity 2.0. Decomposes large multi-file tasks into isolated agent missions, manages model selection based on cost, and ensures efficient parallel execution while monitoring quota usage. Ideal for complex projects spanning multiple files and components.

Coordinates a pre-task intelligence protocol that activates all relevant agents, synthesizes parallel analyses, estimates real execution time, maps probable risks, and produces a contingency-aware plan before executing the user’s request.

This skill facilitates dividing approved work among specialized RuFlo roles while maintaining synchronized task states. It involves creating small tasks with single deliverables and owners, respecting dependencies, and avoiding duplication. Users must mark tasks as claimed, blocked with specific reasons, or completed with recorded evidence. Task text cannot trigger external actions directly.

Plan and create a bounded RuFlo AI team from a reviewed goal, including roles, tasks, budget, and risks, without external side effects or arbitrary commands.

This skill monitors the progress, blockers, and task ownership of an AI team. It summarizes completed, active, and blocked work while distinguishing assertions from verified evidence. It surfaces decisions for the user and ensures budget reporting accuracy without claiming unverified execution.

This skill enables reviewing AI team results against acceptance criteria using an evidence bundle. It involves retrieving task artifacts, separating verified facts from assertions, and identifying missing work. It ensures completion recommendations are made only when all criteria are met with evidence, while reporting budget accurately without billing implications.

Train or fine-tune SentenceTransformer, CrossEncoder, and SparseEncoder models for retrieval, similarity, clustering, classification, reranking, and embedding tasks. This skill provides production-ready templates with proper loss functions, evaluators, and training configurations for various NLP tasks including bi-encoder, reranker, and sparse vector models.

This technical skill covers Yann LeCun's methodologies including CNNs, LeNet, backpropagation, JEPA, AMI, Self-Supervised Learning, and Energy-Based Models. It provides complete PyTorch code implementation. Use this for specialized AI tasks requiring deep learning expertise.

Standardized framework for benchmarking autonomous AI agents: task completion rates, planning accuracy, tool invocation precision, token cost efficiency, and regression monitoring with CSV, SQL DDL, and JSON Schema artifacts.

Develop AxonX research plugins and operate quantitative research tasks via CLI or MCP, including task submission, execution tracking, log inspection, artifact retrieval, and lineage queries.