File Reading
Detect file types and read uploads without binary garbage or context overflow
What You Can Do
When a user uploads a file to Claude, you can intelligently detect its type and read it using the appropriate method. Instead of blindly running cat on binary files or loading massive CSVs into context, this skill routes each file type to the right tool, reading only what's needed to answer the user's question. You'll extract text from PDFs, parse structured data from CSVs and JSON, process images, and decompress archives—all without wasting context or producing garbage output.
Features
inspect extension and MIME type to route correctly
use stat before reading to avoid context overflow on large files
read text content without binary garbage or formatting bloat
load only rows needed, not entire 100MB datasets
extract metadata and visual content when applicable
list contents and decompress selectively without unpacking everything
read and validate without line-wrapping display issues
DOCX, XLSX, ebooks, and plaintext use optimized extraction
Example Output
Example 1: CSV file uploaded
user: I uploaded a sales report. How many rows are there?
clyde: I'll check the file size and row count without loading the whole thing.
$ stat -c '%s bytes' /mnt/user-data/uploads/sales_2024.csv
47283945 bytes
$ wc -l /mnt/user-data/uploads/sales_2024.csv
125847 lines
→ Your CSV has approximately 125,847 rows of data (47 MB).
Example 2: PDF uploaded
user: Can you summarize the key findings in the research paper I uploaded?
clyde: I'll extract the text from the PDF.
$ file /mnt/user-data/uploads/research.pdf
PDF document, version 1.4
[PDF text extraction → first 2000 tokens of readable content]
→ The paper demonstrates that X correlates with Y under conditions Z...
Example 3: ZIP archive uploaded
user: What's in this backup file?
clyde: I'll list the archive contents.
$ unzip -l /mnt/user-data/uploads/backup.zip | head -20
→ Archive contains 342 files: configs/, data/, logs/, reports/...
What's Included
- SKILL.md: core routing logic and protocol for each file type
- File type dispatch table: extension-to-tool mapping for PDFs, CSVs, DOCX, XLSX, JSON, images, archives, ebooks
- Stat & sampling checklist: when to check file size before reading
- Context-safe extraction templates: snippets for each format to avoid overflow
- Fallback protocol: how to handle unknown or corrupted file types
Who It's For
- Data analysts — quickly sample large CSVs and tabular exports without context waste
- Researchers — extract and summarize text from academic PDFs and documents
- Engineers — inspect config files, logs, and structured data in archives
- Product managers — read uploaded reports, spreadsheets, and presentation exports
- Content reviewers — process batches of images, ebooks, and document formats
Best For
- Reading uploaded files without seeing binary garbage or raw ZIP bytes
- Sampling large datasets (CSVs, JSON logs) to answer specific questions quickly
- Extracting text from multi-page PDFs for summarization or analysis
- Inspecting archive contents before deciding what to decompress
- Processing mixed file types in a single upload batch
You might also like
You can systematically identify all licensing requirements for a musical work across multiple territories, coordinate approvals from rights holders and collection societies, and generate legally compliant clearance letters and license terms. This skill maps composition rights (mechanical, performance, synchronization) and sound recording rights (master use, digital performance) while accounting for territory-specific regulations like EU directives, UK post-Brexit rules, and jurisdiction-specific collection society requirements.
by seifbenayed
You can formally verify mathematical proofs by translating theorems into Lean 4 code, leveraging Mathlib's extensive library of formalized mathematics. The skill guides you through researching existing lemmas, designing proof strategies, testing approaches, implementing formal proofs, and verifying correctness—turning informal mathematical claims into machine-checked theorems without requiring deep Lean syntax expertise.
by parcadei
You can read PDFs directly from disk, extract text and embedded content, identify document structure (text-based vs. scanned), list images and attachments, and choose the right extraction strategy based on document type. This skill helps you understand what's in a PDF before processing it, diagnose extraction issues, and pull specific content like tables, forms, and embedded files.
by seifbenayed
You can perform end-to-end PDF operations including extracting text and metadata from documents, merging multiple PDFs into a single file, splitting PDFs by page ranges, filling fillable forms programmatically, and adding text overlays to non-fillable PDFs. This skill handles both simple read operations and complex document workflows, making it essential for document processing, form automation, and PDF batch operations.
by seifbenayed
You can systematically verify licensing compliance across music compositions, calculate accurate mechanical royalties against current statutory rates, and generate audit-ready documentation that satisfies internal controls and external audit demands. This skill consolidates manual spreadsheet workflows into a structured, repeatable process with built-in error detection and cross-referencing against NMPA rates, licensing agreements, and regulatory requirements.
by seifbenayed
You can design complete REST API specifications from requirements, translating them into resource-oriented endpoint architectures that follow industry best practices. Claude generates OpenAPI 3.0 schemas with validated request/response definitions, ensures correct HTTP method semantics and status code usage, and produces Swagger documentation ready for Swagger UI, ReDoc, or Postman integration. Catch design anti-patterns and naming inconsistencies before implementation.
This skill prevents you from providing incorrect information about Anthropic's products by routing you to official documentation. You'll verify details about Claude API (models, rate limits, function calling, batch processing), Claude Code (installation, Node.js requirements, MCP integration), and Claude.ai (subscription tiers, feature limits) before answering user questions. Use it whenever you're uncertain about product specs, pricing, or capabilities.
by seifbenayed
Query your local Screenpipe instance to retrieve screen recordings, audio transcriptions, UI element accessibility trees, keyboard/mouse input logs, and productivity analytics. You can search by keywords, filter by content type (audio, OCR, accessibility, input), set time ranges, and extract structured data about your applications, meetings, and work sessions without sending data to external servers.
by az9713