AI Engineering

Brainstorm runs a compact diverge-to-converge loop that helps users generate, challenge, compare, and narrow ideas. Use it for ideation, evaluating options, naming, technical approaches, content angles, or strategic decisions.

Dev Lifecycle gives coding agents a repeatable engineering workflow that moves from requirements and design through planning, implementation, testing, and review. Use it to make Claude Code or Codex work like a senior engineer instead of jumping straight into code.

Dev Requirements guides the capture and review of requirements for AI-assisted development, including clarifying scope, initializing feature docs, and validating requirements before design. Use it when starting a new feature or reviewing existing requirements in the dev-lifecycle phases 1-2.

Memory CLI stores verified, reusable project knowledge for AI coding agents and searches it before non-trivial work. Use it to persist conventions, fixes, and decisions while avoiding secrets and task progress.

LangGraph provides guidance on implementing the skills pattern for AI agents, where specialized prompts and tools are loaded on-demand for progressive disclosure. Use this skill when building multi-agent systems with LangGraph/LangChain that need modular, team-distributed capabilities.

Hivemind stores and retrieves global team memory across sessions, users, and agents in an Activeloop organization. Use it to recall shared context, search summaries, and sync mined Claude skills with a team.

Hivemind Graph queries a local AST-derived code graph of functions, classes, calls, and imports to answer structural questions like what calls X or where Y is defined. Use it to map dependencies and blast radius before changing a signature.

Hivemind Memory provides global team and org memory for AI agents, searching both built-in and shared Hivemind memory when recalling information. Use it to look up past sessions, manage organization access, share reusable skills, and run semantic memory searches.

Hivemind Memory stores persistent team and organization memory in ~/.deeplake/memory/, including session summaries and raw logs. Use it when you need to recall information from past sessions or share reusable skills across a team.

Wiki Capture records durable business knowledge like metric definitions and organizational conventions so an analytics assistant can reuse them across chats. Use it to capture user preferences, rules, and mappings that are not already in the semantic layer.

Sandbox Agent provides a universal API for orchestrating AI coding agents (Claude Code, Codex, OpenCode, Amp) in sandboxed environments. Use it to deploy and configure the server, manage agent sessions via SDK or API, stream events, and build chat UIs for coding agents.

Ad Hoc Profile Load is a legacy alias for the canonical spk-doctrine-profile-load skill. It resolves a requested profile and applies action-scoped governance, specialization boundaries, directives, and collaboration handoffs when an agent needs to adopt a role such as architect, reviewer, researcher, or planner.