Glossary: Domain 2 — File Management, Data Processing & Output

Direct Local File Access

Cowork's ability to read, write, edit, and create files directly on your local file system without any upload or download step. Unlike Chat where you drag files in one at a time (limited to 30MB per file and 20 files per conversation), Cowork reads files directly from disk with no size constraints on individual files and processes entire folders at once.

Exam context: Questions test whether you know the difference between Chat's upload model and Cowork's direct access. The 30MB limit applies to Chat only — not to Cowork.

Working Folder

The root directory of Claude's world for a given session. Everything Claude reads, creates, and modifies lives within this boundary. You select the working folder when starting a session, and Claude cannot access files outside it. Best practice is to create a dedicated, purpose-built folder for each task rather than pointing at broad directories.

Exam context: Know that folder access is granted per session (not globally), that each new session requires selecting a working folder, and that recently used folders are remembered for quick selection.

Dedicated Workspace

A purpose-built folder containing only the files you want Claude to work with, following the principle of least privilege. Instead of pointing Cowork at your entire Documents folder, you create a folder like "Q2-Expenses-Processing" and copy only the relevant files into it. This limits the blast radius of any security incident.

Exam context: The "dedicated workspace" best practice appears in nearly every security question. The correct answer almost always involves creating a scoped folder rather than granting broad access.

Cross-Document Synthesis

Cowork's ability to read multiple documents simultaneously, extract insights from each, and produce a unified analysis that identifies patterns, contradictions, and themes across the full set. Sub-agents can process different documents in parallel, making synthesis of 10+ sources practical in minutes rather than hours.

Exam context: Know that all documents must be in the same working folder for cross-referencing to work, that sub-agents work independently during extraction and the main agent synthesises afterwards, and that specifying the output format is critical.

Structured Data Extraction

The process of reading an unstructured source (such as a scanned invoice or free-form report) and producing organised, tabular output (dates, amounts, categories, names) written directly to a spreadsheet. The pipeline is: read unstructured source, interpret content, write structured output — with no intermediate copy-paste step.

Exam context: Questions may present a scenario with unstructured PDFs and ask for the best approach to extract specific data points into a structured format.

Working Formulas

Excel formulas (VLOOKUP, SUMIFS, INDEX-MATCH, IF) that Cowork writes as functional, recalculating formulas — not static values that merely look like formulas. When you change a value in the source data, the formulas update automatically. This is a key differentiator from basic data dumps.

Exam context: A common exam trap claims that Cowork-generated Excel files contain static values. They do not — Cowork writes real, working formulas that recalculate when data changes.

Conditional Formatting

Rules applied to Excel cells that automatically change their visual appearance based on the data they contain. Cowork can apply conditional formatting — for example, highlighting any expense over $500 in red or flagging regions below target in amber — as part of a workbook creation task.

Exam context: Know that conditional formatting is part of Cowork's Excel capabilities but requires explicit specification in your prompt. Without instructions, Claude defaults to the simplest output.

Direct File System Delivery

In Cowork, files are written directly to your working folder on completion — no download step required. When Claude finishes creating your Excel report, it is already sitting on your file system, immediately available to Finder, email clients, and cloud sync services. This contrasts with Chat, where outputs appear as download links.

Exam context: Questions test whether you understand the difference between Chat delivery (download links) and Cowork delivery (direct to file system). Know that Cowork's direct delivery eliminates manual steps.

Batch Processing

The processing of multiple files in a single Cowork task, typically with parallel sub-agents. Phrases like "for each one," "analyse all," and "process these" signal to Claude that the work can be parallelised. Batch processing is ideal for renaming, converting, extracting data from, or summarising many files at once.

Exam context: Know the natural language triggers that activate parallel batch processing and when batch processing is appropriate versus when sequential processing is needed.

Data Transformation

Converting data from one format or structure to another — for example, merging 50 CSVs into a single Excel workbook with pivot tables, or converting raw JSON exports into formatted PDF reports. Cowork handles the entire pipeline: reading source files, applying transformation logic, and writing the output.

Exam context: Questions present transformation scenarios and test whether you can identify the correct Cowork approach versus manual alternatives or Chat-based workarounds.

Output Format Specification

The practice of explicitly describing how you want analysis or documents delivered — including file type, naming convention, column names, chart types, tab structure, and section headings. "Create a spreadsheet" produces a flat data dump; specifying formulas, charts, and conditional formatting produces a professional tool.

Exam context: Nearly every document creation question rewards the answer with the most specific output format specification. The quality ceiling is high, but you must raise it through your prompt.

File Organisation Workflow

An automated process for sorting, renaming, and restructuring files within a working folder according to specified rules. Common patterns include organising by date, type, or project; applying naming conventions; archiving old files; and generating inventory manifests of changes made.

Exam context: Questions test whether your proposed workflow includes clear naming conventions, archive rules, and explicit guardrails (such as "do not delete anything without permission").

See also: 2.5 File Organisation & Cleanup

Permission Tiers

The three levels of the Cowork permission model: (1) folder scoping — Claude can only access what you grant; (2) operation type — reads and writes happen freely but deletions require explicit approval; (3) Computer Use — accessing applications outside the VM requires separate per-application permissions.

Exam context: Understanding all three tiers is essential. Many candidates only know about deletion protection and miss the distinction between in-VM operations and Computer Use permissions.

Least-Privilege Scoping

The security principle of granting Claude access to only the minimum set of files and folders needed for the task. Never point Cowork at broad directories like Documents, Desktop, or your home folder. Create a dedicated working folder and copy only the relevant files into it.

Exam context: This principle underpins the correct answer to virtually every security-related file access question.

Multi-Format Analysis

Cowork's ability to read and analyse different file types — PDFs, XLSX, PPTX, CSV, text documents, and images — within a single session from the same working folder. This allows combined qualitative and quantitative analysis across diverse source materials.

Exam context: A trap answer may claim Cowork can only analyse text-based documents. It can read PDFs, spreadsheets, presentations, and many other formats directly from the file system.

Glossary: File Management, Data Processing & Output

Terms in this domain