deep-dive

Detects your project's full technology stack, then generates a comprehensive research prompt tailored to a specific topic like security, performance, or testing.

Using FilesystemMiddleware with virtual filesystems, backends (State, Store, Filesystem, Composite), and context management for Deep Agents.

Run a comprehensive deep review combining architecture analysis, code review, error handling audit, type design analysis, comment verification, test coverage analysis, accessibility audit, localization review, concurrency analysis, performance analysis, code simplification, and platform-specific reviews (iOS, macOS, Android, Angular, TypeScript, Next.js, Vue.js, Python, Django, Ruby, Rust, Go, Rails, Flutter, Java/Spring Boot, C#/.NET, PHP/Laravel, C/C++, React Native, Svelte/SvelteKit, Elixir/Phoenix, Kotlin Server, Scala, Terraform, Shell/Bash, Docker, Kubernetes, GraphQL, GitHub Actions, SQL, Swift Data). Platform reviewers are automatically included when relevant. Distinguishes between NEW issues (introduced by PR) and PRE-EXISTING issues (technical debt). Use when reviewing PR changes, before merging, or for thorough code quality assessment. Supports flags --pr, --branch, --changes for scope detection.

Multi-agent deep research pipeline for complex questions (EIP analysis, architecture decisions, cross-client comparisons, protocol design). Use when single-shot answers are insufficient and you need decomposition, parallel investigation, adversarial critique, and a formal output document.

Use when reading books, academic papers, or long-form documents from URLs or PDFs and needing structured analysis, comprehension, or comparison

Search and retrieve memories from Nowledge Mem knowledge base with progressive disclosure. This skill should be used when the user asks to search memories, recall past knowledge, find saved information, look up conversation history, expand thread details, or mentions keywords like "记忆", "知识库", "之前说过", "我保存的", "历史对话".

DEPRECATED - Use gpt-researcher skill instead. This browser automation approach has been replaced with the local GPT-Researcher framework using OpenAI o1.

Write Python code in the style of David Beazley, author of Python Cookbook. Emphasizes generators, coroutines, metaprogramming, and understanding Python's internals. Use when writing advanced Python that requires deep language mastery.

Enforce Design-It-Twice workflow: generate 2-3 radically different approaches, compare them, then implement. Use when designing modules, APIs, or classes before implementation. Triggers on: design, create class, add module, implement feature, new service, API design, before implementing. Produces structured design document with approaches, comparison table, choice rationale, and depth check.

Performs deep structured analysis on complex or ambiguous problems. Activates when problems are unclear, have multiple perspectives, or require careful thinking before proceeding. Uses ultrathink methodology for systematic exploration of problem space.

Transparent, rigorous research with full methodology — not a black-box API wrapper. Conducts exhaustive investigation through mandated 2-cycle research per theme, APA 7th citations, evidence hierarchy, and 3 user checkpoints. Self-contained using native OpenClaw tools (web_search, web_fetch, sessions_spawn). Use for literature reviews, competitive intelligence, or any research requiring academic rigor and reproducibility.

[22] UNDERSTAND. Deep research from all sources — internal (code, git, logs) AND external (web, docs, best practices). Use when choosing libraries, investigating solutions, understanding legal/technical questions, comparing approaches, or anytime you need comprehensive knowledge from both project context and world knowledge. Triggers on "research", "investigate", "find best approach", "what do others do", "compare options", or complex questions requiring multiple sources.