Document Skills

# Document Authoring Skill Use this skill whenever you need to **create, compose, or substantially revise** a document. It bridges output formatting (how to write markdown) with document management (how to store it) and gives you a clear workflow for documents of any size. **Tone:** Write professionally. Do not use emojis in document content — they look unprofessional in exported PDFs and formal documents. > For pure CRUD operations (rename, delete, move to collection), call `learn_document_management`. > For formatting reference (chart syntax, fusion fragments, expand blocks), call `learn_output_reference`. --- ## 1. Plan Before You Write Always outline before generating content. The cost of re-writing a badly structured document is much higher than spending one turn on planning. ### Workflow 1. **Clarify scope with the user** — ask about audience, length, tone, and any source documents. 2. **Propose an outline** — use markdown headings. Include estimated section lengths if t...

# Conversation Analysis Skill Use this skill to find past agent conversations and analyze them with AI. ## Search Conversations Use `search_conversations` to find past workflow runs: ``` // List recent conversations search_conversations({}) // Filter by status search_conversations({ status: "COMPLETED", page_size: 20 }) // Filter by time range search_conversations({ start: "2024-01-01T00:00:00Z", end: "2024-01-31T23:59:59Z" }) // Filter by interaction name search_conversations({ interaction: "MyInteraction" }) ``` Results include `workflow_run_id` and `workflow_id` needed for analysis. ### Pagination When more results exist, the response includes `has_more: true` and a `next_page_token`. Pass the token to get the next page: ``` search_conversations({ status: "COMPLETED", next_page_token: "token_from_previous" }) ``` ## Analyze a Conversation Use `analyze_conversation` to perform AI-powered analysis on a conversation: ``` // Summarize what happened analyze_conversatio...

# Custom PDF Templates Use this skill when the user wants to **customize the look of PDF exports** — branding, colors, fonts, header/footer layout, title page design, or logo replacement. The `render_markdown` tool accepts `template_path` and `logo_path` parameters that let you supply a custom LaTeX template and logo stored as artifacts. --- ## 1. Quick Start ``` 1. write_artifact(path: "files/my-template.latex", content: "<modified template>", type: "file") 2. write_artifact(path: "files/logo.png", source_url: "https://...", type: "file") // optional 3. render_markdown( artifact_path: "files/report.md", template_path: "files/my-template.latex", logo_path: "files/logo.png", title: "Branded Report" ) 4. inspect_pdf(pdf_path: "out/Branded Report.pdf") ``` --- ## 2. Protocol URI Formats Both `template_path` and `logo_path` accept: | Format | Example | Resolution | |--------|---------|------------| | Bare artifact path | `fil...

# Document Management Skill Use this skill to manage documents and collections in the content store. This covers CRUD operations, importing content, and organizing documents into collections. ## Document Operations ### Fetch a Document Use `fetch_document` to retrieve a document by ID. Supports multiple modes: | Mode | Description | |------|-------------| | `full` | Complete document with all metadata | | `properties` | Metadata, custom properties, and `headings` array (for markdown docs: level, title, line number) | | `content` | Only the text content | | `sections` | Document split into sections | | `analyze_full` | AI-powered analysis of the document | | `analyze_sections` | AI analysis of specific sections | | `analyze_range` | AI analysis of a token range | Optional parameters for content filtering: - `page_range`: `{start, end?}` — filter by page markers (PDFs). 1-indexed, inclusive. - `line_range`: `{start, end?}` — filter by line numbers. 1-indexed, inclusive. Use with `...

# Document Search Skill Use this skill to search and query documents in the content store. Two tools are available: | Tool | Best For | |------|----------| | `query_documents` | SQL queries, aggregations, analytics, complex filters | | `search_documents` | Simple searches, filters by status/type/collection | ## Quick Search with search_documents For simple searches, use `search_documents`: ```json { "query": { "full_text": "quarterly report" } } { "query": { "status": "published" } } { "query": { "type": "<type_id>" } } { "query": {}, "collection_id": "<collection_id>" } { "query": { "full_text": "budget", "status": "published", "type": "<type_id>" }, "limit": 20 } ``` ## Advanced Queries with query_documents For complex queries, aggregations, and analytics, use `query_documents` with SQL, ES|QL, or DSL. ### SQL Queries (Recommended) Standard SQL syntax. Best for filtering, sorting, and basic aggregations. ```sql -- Find published documents SELECT name, status, created_at FR...

# Presentation Authoring Create professional slide decks rendered as PDF using the `render_slides` tool. Write natural markdown with `---` slide separators — templates are auto-detected from content structure. ## Workflow 1. **Plan** — Outline the deck structure (number of slides, key messages) 2. **Compose** — Write the slide markdown with frontmatter 3. **Render** — Call `render_slides(markdown: "...")` to produce the PDF 4. **Inspect** — Call `inspect_pdf` to visually review the output 5. **Iterate** — Fix issues and re-render until the user is satisfied ## Markdown Format ```markdown --- title: Q4 Business Review font: Inter backgroundColor: "#ffffff" --- # Q4 Business Review ## Annual Performance Summary --- ## Key Highlights - Revenue up 15% year over year - Expanded to 3 new markets - Team grew from 35 to 50 people --- ## Financial Details --- ## Revenue vs Cost :::left - SaaS: $2.1M - Services: $800K - Consulting: $300K ::: :::right - Infrastructure: $500K - Hir...

# Content Summarization Skill When summarizing content, follow these structured approaches: ## Summarization Strategies ### 1. Executive Summary (For Business Documents) - **Length**: 1-2 paragraphs - **Structure**: Key decision/recommendation → Supporting points → Action items - **Focus**: What matters to decision-makers ### 2. Abstract Style (For Technical/Academic) - **Length**: 150-300 words - **Structure**: Purpose → Methods → Key Findings → Conclusion - **Focus**: Reproducibility and scientific accuracy ### 3. TL;DR Style (For Casual/Quick Reference) - **Length**: 2-4 sentences - **Structure**: Main point → Why it matters - **Focus**: The single most important takeaway ### 4. Bullet Point Summary - **Length**: 5-10 bullet points - **Structure**: Key points in order of importance - **Focus**: Scannable, actionable items ## Best Practices 1. **Preserve key information**: Names, dates, numbers, and specific claims 2. **Maintain original meaning**: Don't introduce interpreta...

# Document Translation Skill Use this skill when translating documents between languages. It covers strategies for maintaining quality and consistency, especially for long documents where section-by-section translation is required. > For general document composition workflows, call `learn_content_authoring`. --- ## 1. Before You Start ### Clarify with the user 1. **Target language** — confirm the exact language and regional variant (e.g., French (France) vs. French (Canada), Brazilian Portuguese vs. European Portuguese) 2. **Tone and register** — formal (vous/usted), informal (tu/tú), technical, legal, marketing? 3. **Terminology preferences** — does the user have preferred translations for domain-specific terms? Are some terms left untranslated (brand names, product names, technical acronyms)? 4. **What to preserve** — code blocks, URLs, proper nouns, formulas typically stay untranslated 5. **Existing translation** — does a translated version of this document already exist? If ...

# Zip Archive Handling Skill Use this skill whenever you need to work with **zipfiles** stored in Vertesia (for example, zipped collections of CSVs, logs, or documents). Key behaviors: - A Vertesia **object** whose source is a `.zip` file has a **text rendition** that lists the files inside the archive (an index), not their full contents. - To actually read files from the archive, you must download the zip into the Daytona sandbox and unzip it there. --- ## 1. Understand the Text Rendition When you fetch or inspect a zip object’s text rendition (for example via `fetch_document` in analyze mode), it typically contains: - A list of file paths and names inside the archive. - Sometimes sizes and timestamps, depending on how the index was generated. Use this index to: - Identify which internal files matter (e.g. `data/metrics.csv`, `reports/summary.md`). - Decide which ones you want to extract and analyze. Do **not** treat the text rendition as the full content of each file—it is...