AI Engineering

Do gives Claude Code persistent long-term memory by storing and retrieving embeddings in SQLite and vector databases. Use it when you want the agent to remember context across sessions.

OpenClaw Memory sets up persistent memory for agents running on an OpenClaw gateway by injecting an observation timeline into system prompts through the claude-mem plugin. It also offers an optional real-time observation feed to a messaging channel.

Agent Skill Discovery discovers and invokes agent skills, helping an agent identify the right engineering workflow for a task. Use it when starting a session or when unsure which skill applies.

Idea Refine turns vague ideas into sharp, actionable concepts through structured divergent and convergent thinking. Use it when an idea is still unclear, before committing to a plan, or to stress-test assumptions.

Interview Me asks one question at a time with attached guesses to uncover what a user actually wants instead of what they think they should want. Use it when an ask is underspecified, conventional, or the user explicitly asks to be interviewed before planning or building.

Planning and Task Breakdown decomposes work into ordered, verifiable tasks with explicit acceptance criteria. Use it when you have a spec or clear requirements, when a task feels too large to start, or when parallel work is possible.

Spec-Driven Development writes a structured specification before coding, defining objective, commands, code style, testing strategy, and boundaries. Use it when starting a new project or feature, when requirements are ambiguous, or when a change touches multiple files.

Context Engineering optimizes the information an AI agent sees by structuring rules files, specs, source files, error output, and conversation history. Use when starting a session, when agent output degrades, or when setting up project context.

Audio Jingle routes music, speech, and sound-effect requests to the configured audio models and saves one MP3 or WAV file in the project. Use it to generate jingles, voiceovers, beds, or SFX from project metadata.

Understand Anything analyzes a codebase to produce an interactive knowledge graph for exploring architecture, components, and relationships. Use it to map dependencies and document project structure.

Understand Chat answers questions about a codebase using a pre-built knowledge graph, so an agent can explain files, functions, and dependencies. Use it when you need to ask questions about a codebase or understand code without reading the entire graph into context.

Understand & Explain provides a thorough, in-depth explanation of a specific file, function, or module in a codebase by combining a knowledge graph with direct source code reading. Use it when you need to understand how a component works, its architecture layer, and its connections.