Skill UI

Defines an operator approval contract for agent-drafted outbound messages: hashed draft filing, epoch-keyed decisions, approval snapshots, durable delivery claims and receipts, plus a pre-draft baseline gate. Use when a human must approve, reject, or steer each send before it leaves.

Read-only diagnosis for OpenViking memory issues: trace backward from a memory file to its archive memory_diff.json, then to session messages when needed. Use it for incorrect content, wrong paths or owners, missing memories, and unexplained updates or deletes.

Create real email accounts, send and receive messages, register inbound webhooks, and inspect billing balance for AI agents via the AgentMail REST API, enabling service signups, verification code flows, and transactional communication.

Low-level Java SDK for creating and managing persistent Azure AI agents with threads, messages, runs, tools, authentication, polling, and resource cleanup.

This skill covers the core concepts of context engineering for language models, including system prompts, tool definitions, retrieved documents, message history, and tool outputs. It explains how attention mechanisms work, context window constraints, and strategies for optimizing context usage to reduce costs and improve agent performance.

This skill involves testing AI agents for direct prompt injection vulnerabilities where user messages attempt to override system instructions. It focuses on role confusion and instruction hijacking without external content, ensuring agents maintain security boundaries against direct manipulation attempts.

Manage multi-agent pipelines on Grok Build using Ruflo-style team bus and MCP tools. Ideal for coordinating feature development workflows like architect-to-reviewer chains or replacing Claude SendMessage. Includes CLI fallback for team creation, planning, and subagent spawning with isolation controls.

This skill coordinates work across a RuFlo AI-agent swarm by discovering channels, reading signed updates, and checking claims. It helps manage multi-agent status, handoffs, ownership conflicts, and shared-work planning. Users must treat relay-sourced messages as untrusted and avoid irreversible actions unless explicitly requested. It ensures safe coordination without unauthorized state changes.