> For the complete documentation index, see [llms.txt](https://docs.diaflow.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.diaflow.io/chat/skills.md). # Skills ## What is skill? In the Diaflow framework, a skill is a modular, reusable package of specialized workflows, domain guidelines, and execution templates that expands an autonomous AI agent's core capabilities. Rather than relying on generic instructions for every request, skills equip the agent with structured operational procedures, standard operating patterns, and tool configurations tailored for specific outcomes (ranging from document parsing and interactive data visualization to programmatic video creation). By encapsulating domain expertise into standardized modules, skills allow agents to execute complex, multi-step tasks with consistent precision, operational autonomy, and high output quality. ## How to use skills Type `/` in the input bar and type in the skill name (i.e. `/slide-generation`) or simply mention the skill name to use it (i.e. Use slide generation to generate a presentation). ## Creating skills Diaflow allows you to create and upload custom skills to tailor your AI agents to company guidelines, internal operating procedures, or specialized execution workflows. You can upload a skill as a standalone Markdown file or as a packaged ZIP archive. ### File requirements When uploading a skill through the **Upload a skill** modal, your file must meet one of the following requirements: * **Single Markdown file (`.md`):** Must contain a YAML frontmatter header with both a `name` and a `description`. This format is ideal for prompt-based workflows, style guides, decision rubrics, or standard operating procedures (SOPs). * **ZIP archive (`.zip`):** Must include a `SKILL.md` file at the root level of the archive. This format is recommended when bundling supporting scripts, reference data, rule files, or templates. ### Skill specification format Every custom skill is defined by a markdown file (`SKILL.md` in a ZIP archive, or the standalone `.md` file) composed of two core sections: #### 1. YAML frontmatter The file must start with a YAML frontmatter block enclosed between triple dashes (`---`): ```yaml --- name: your-skill-name description: A detailed description of what the skill does and specific triggers for when the AI agent should use it. --- ``` * **`name`:** A unique, lowercase, hyphen-separated identifier (for example: `brand-voice-checker` or `meeting-notes-synthesizer`). * **`description`:** A clear, informative summary explaining the skill's capabilities and exact conditions under which the agent should activate it. The AI agent relies on this description to automatically determine task relevance. #### 2. Instruction body The markdown body contains the operational knowledge and procedures for the AI agent to follow: * **Role & Objective:** What role the agent assumes and the primary goal of the skill. * **Core Guidelines:** Specific rules, rubrics, tone standards, or decision logic. * **Workflow Steps:** Step-by-step instructions detailing how the agent should process inputs. * **Output Format:** Expected output structure (for example: executive summary, bulleted action items, or structured markdown tables). ### Directory layout for packaged skills For skills that include custom helper scripts, references, or assets, organize your files into a clean structure before compressing them into a `.zip` archive: ``` my-custom-skill/ ├── SKILL.md # Required entrypoint with YAML frontmatter ├── scripts/ # Optional helper scripts or CLI utilities │ ├── run.py │ └── requirements.txt ├── references/ # Optional reference documents, rubrics, and guidelines │ └── brand-rules.json └── assets/ # Optional starter templates or sample files └── report-template.docx ``` ### Example: Brand Voice Checker Here is a practical example of a complete `SKILL.md` file designed for reviewing marketing and website copy against brand guidelines: ```markdown --- name: brand-voice-checker description: Reviews and refines marketing copy, email campaigns, and product documentation against official brand voice guidelines. Use when the user asks to check tone of voice, eliminate buzzwords, or ensure consistent messaging. --- # Brand Voice Checker ## Role & Goal You are an expert brand editor. Your role is to analyze draft copy and ensure it aligns with corporate messaging standards, maintains an active tone, and avoids prohibited buzzwords. ## Core Guidelines 1. Tone Pillars: - Clear and concise: Prioritize simple language over complex jargon. - Confident and helpful: Speak directly to the reader using second-person perspective ("you", "your"). - Action-oriented: Use active voice and strong verbs. 2. Prohibited Words and Alternatives: - Avoid "synergy" (use "collaboration"). - Avoid "leverage" as a verb (use "use" or "apply"). - Avoid "game-changing" (use "impactful" or describe specific benefits). ## Operational Workflow 1. Read and analyze the user's provided text. 2. Evaluate adherence to the three tone pillars. 3. Identify any prohibited words or passive voice constructions. 4. Provide a structured review with concrete rewrite suggestions. ## Output Format Format your response in three clear sections: ### 1. Overall Assessment A brief score (1 to 10) and summary of how well the draft meets brand standards. ### 2. Line-by-Line Suggestions A table with three columns: - Original Text - Suggested Revision - Reason for Change ### 3. Polished Version The complete revised copy ready for publishing. ``` ### Uploading a skill in Diaflow 1. Open **Diaflow Chat**. 2. Click the skill management icon in the input area and select **Upload a skill**. 3. Drag and drop your `.md` file or `.zip` archive into the upload window. 4. Diaflow verifies that the YAML frontmatter and `SKILL.md` satisfy the required specifications. 5. Click **Upload & Preview** to review the parsed metadata and activate the skill for your workspace. ## Default skills | Skill name | Description | | ------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `slide-generation` | Creates high-design HTML & PowerPoint presentation decks, pitch decks, and keynotes across distinct aesthetic styles. | | `html-to-video` | Converts handwritten HTML, CSS, and JavaScript into production-ready MP4 videos or PNG/JPEG images via headless Chromium (Playwright) and FFmpeg. Ideal for motion graphics, dynamic headers, data visualization clips, and social cards. | | `motion-web-builder` | Builds production-grade, motion-driven websites and landing pages from user briefs, screenshots, videos, or URLs. Implements cinematic web experiences with React, Next.js, Vite, and Tailwind CSS. | | `diaflow-execution-blueprint` | Transforms any idea, project, or business initiative into a production-ready Diaflow guidebook, orchestrating Kiwi, Drive, Chat, Teammates, Pages, Tables, and Studio. | | `systematic-literature-review` | Conducts systematic literature reviews, surveys, and multi-paper syntheses on specific research topics. Searches arXiv and exports comprehensive reports in APA, IEEE, or BibTeX formats. | | `academic-paper-review` | Reviews, analyzes, critiques, and summarizes academic papers, preprints, and scientific publications. Delivers structured methodology assessments and constructive academic feedback. | | `data-analysis` | Performs comprehensive data analysis, statistical summaries, pivot tables, and SQL queries on uploaded Excel (.xlsx/.xls) and CSV datasets. | | `consulting-analysis` | Generates professional management consulting reports including market analysis, consumer insights, brand strategy, financial assessments, and strategic frameworks. | | `surprise-me` | Delivers delightful and creative "wow" experiences by proactively exploring and creatively combining multiple enabled Diaflow skills and capabilities. | | `podcast-generation` | Transforms written articles and documents into engaging, conversational podcast audio scripts featuring natural two-host dialogue. | | `newsletter-generation` | Researches, writes, and formats professional email newsletters, weekly executive digests, industry roundups, and curated content summaries. | | `github-deep-research` | Conducts deep, multi-turn investigations into GitHub repositories, delivering comprehensive architectural breakdowns, timeline reconstructions, and competitive open-source analyses. | | `code-documentation` | Generates and enhances technical documentation, including READMEs, API references, inline code comments, system architecture guides, and changelogs. | | `bootstrap` | Creates a personalized SOUL.md identity file through an adaptive onboarding conversation to initialize and configure custom AI partner personas. |