OfficePlane

Generate a Word document from ingested source documents using the agnostic Document tree (CommonMark/Pandoc-aligned)

Generate a single Document (Word/PowerPoint) from all source documents in a collection, with attributions back to each source

Generate a PDF from ingested source documents using the agnostic Document tree

Generate a PowerPoint deck from ingested source documents using the agnostic Document tree (parametrised by slide_count, style, audience, tone)

Generate an Excel workbook (.xlsx) from ingested source documents — tabular with charts, formulas, KPI cells

Find PII (emails, phones, SSN, credit cards, IBANs, names, addresses, DOB) in an ingested document; return redaction plan

Extract tabular data from one or more source documents and populate a saved xlsx template — composes extract-tabular-data + xlsx-template-apply

Walk an ingested document's pages and extract every detected table into a canonical {name, headers, rows} structure with source attributions

Apply a saved xlsx template with new data — fills each table by name, renders .xlsx

Save a generated workbook's shape (headers, formats, charts, KPIs) as a reusable template — strips data rows

Analyze an ingested .xlsx document for formula errors, suspected typos, outliers, missing totals, and dead references

Rewrite a single node in a generated document using DeepSeek, given a user instruction and surrounding context. Does NOT mutate the document — caller applies the new node via document-edit replace.

Validate that each attribution in a generated document is actually supported by the cited source span; flag unsupported claims

Apply insert/replace/delete operations to a workspace document.json by node_id

Suggest which existing Collection an ingested document belongs to, and propose a new Collection name if none fit

Semantic search across ingested documents. Returns top-k passages.

Linux environment reference — pre-installed tools (LibreOffice, Chromium, Node.js, pptxgenjs, python-pptx, matplotlib), useful commands, and packaging workflow.

Block-based course content generation — schema v1.0 (modules → lessons → blocks). Use when writing course_content.json with text/title/table/image blocks grouped by lesson.

Create distinctive HTML slide presentations with reveal.js. Custom themes, Google Fonts, Playwright screenshots for verification.

Generate educational illustrations via Google Gemini. Use the pre-installed /app/scripts/generate_image.py — never draw with matplotlib or PIL.ImageDraw for illustrations.

Create polished PPTX presentations with pptxgenjs. Includes color palettes, typography rules, layout patterns, QA workflow, and rendering commands.

Read-only query over the ExecutionHistory table with structured filters.

Create folders, move documents between folders, and manage ACL inheritance.

Apply a taxonomy label, sensitivity, and retention bucket to a document.

Render a document to a requested output format and return a download path.

Extract structured fields from a document using a caller-supplied JSON schema.

Upload, vision-parse, and persist a document with multi-level summaries.

Strip PII and sensitive entities from a document and store a redacted derivative.

Create or query directed relations between documents as graph edges.

Hybrid vector + full-text search across the document store with hierarchical context.

Generate or refresh document, chapter, and section summaries top-down.

Create a new document version, diff against the prior version, or roll back to a previous version.

Drive an approval or sign-off state machine on a document through defined workflow actions.

Test fixture skill — do not invoke in production

Design and build Python systems using Domain-Driven Design (DDD), Clean Architecture, and SOLID principles. Uses FastAPI and SQLAlchemy.

The full develop-validate-test-review-improve loop. Use as the primary workflow after implementing any feature or fix.

Run an entropy sweep to detect and fix code drift — duplication, naming inconsistencies, dead code, pattern divergence, file bloat, and doc content staleness. Use weekly or after major features.

Design and structure React applications using Feature-Sliced Design (FSD). Enforces strict unidirectional data flow, layer isolation, and business-value slicing.

Improve and maintain the agent harness — hooks, checks, skills, agents, contracts, enforcement architecture. Use when adding rules, fixing enforcement gaps, or auditing harness compliance.

Run all 5 specialized reviewers on recent changes and aggregate results.