Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Sci-EDA Context Shield for Scientific Data

    1

    Use when inspecting CSV, Parquet, Jupyter notebooks, NumPy, Safetensors, HDF5, Zarr, MATLAB, or WAV datasets in one shot without overflowing agent context.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Sci-EDA Context Shield for Scientific Data

    Sci-EDA Context Shield for Scientific Data

    Example session with this skill installed

    Inspect all datasets, model weights, audio clips, and notebooks inside ./data_release/ in one pass and summarize their schemas, tensor shapes, and potential context hazards.

    • Read your context and instructions
    • Compiled the sci-eda context shield
    • Generated the data export

    Ran python3 scripts/sci_eda_probe.py ./data_release --json (completed in 0.18s, output 2.4 KiB vs. 614 MiB raw files on disk):

    FileSizeFormat / EngineKey Schema, Shapes & Diagnostics
    encoder_fp16.safetensors164.2 MiBsafetensors (stdlib header)42 tensors, 86,018,048 params (F16); layers.0.attn.qkv.weight: [2304, 768]
    cohort_expr.csv412.0 MiBcsv (stdlib stream)18,420 rows × 2,048 cols; HAZARD: single line width 19.4 KiB (>8 KiB wide matrix); 3.2% missing in batch_cov
    ablation_plots.ipynb28.6 MiBipynb (stdlib AST/JSON)34 cells (22 code, 12 md), kernel Python 3.11; 97.8% size is embedded base64 PNGs (stripped safely)
    hydrophone_ch01.wav9.2 MiBwav (stdlib wave)48,000 Hz, 1 ch, 16-bit PCM, duration 96.000s

    Actionable Findings:

    1. Do

    not call raw text Read on cohort_expr.csv or ablation_plots.ipynb; use lazy column projection (polars.scan_csv on target gene columns).
    2. encoder_fp16.safetensors hidden dimension is confirmed 768 (F16), matching configs/encoder_base.yaml.

    cohort_expr.csv

    CSV · data export

    Generated

    Example file from a real run - the skill writes it into your workspace.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    One accidental Read on a 300 MB .csv, a wide genomics matrix, or a base64-bloated .ipynb notebook can burn $3+ in tokens and corrupt a 2-hour coding session instantly.

    Sci-EDA Context Shield ships a 100% Python 3.9+ stdlib CLI (scripts/sci_eda_probe.py) + agent workflow that inspects 15+ scientific, ML, and audio formats in a single round-trip with a hard <6 KiB stdout cap:

    • Zero-Dependency Stdlib Parsers Out of the Box:
      • .safetensors: Parses the 8-byte little-endian header directly to extract exact tensor names, shapes, dtypes, and total parameter counts—even without torch or safetensors installed!
      • .npy / .npz: Reads binary NumPy headers natively (upgrading automatically to numpy mmap stats if installed).
      • .ipynb: Strips giant base64 image outputs, flags bloat warnings, and prints a clean cell TOC + code preview.
      • .csv / .tsv / .vcf: Streams row/column counts, missingness rates, and detects ultra-wide (>8 KiB) matrix rows.
      • .wav / .aiff, .zarr directory trees, .json / .jsonl, and .gz archives.
    • Automatic Scientific Stack Enrichment: Seamlessly activates polars/pandas (.parquet/.feather), h5py/scipy (.h5ad/.h5/.mat), and torch (.pt/.ckpt) when present in your environment.
    • Dual Local + Remote MCP Ready: Works out of the box across Claude Code, Cursor, Copilot, Windsurf, Codex, and Google Antigravity.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.

    Does not load entire tensor arrays into memory by design. For deep numerical computing on giant arrays, use polars, pyarrow, or dask after inspecting schemas with this skill.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 2 days ago

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    This skill prevents agents from crashing or hitting token limits by providing a 6 KiB-capped summary of large scientific files. It extracts critical metadata, shapes, and schemas without loading the entire dataset into the agent's memory.

    The skill natively supports over 15 scientific formats including Safetensors, Zarr, HDF5, Parquet, and NumPy. It also handles Jupyter notebooks by stripping out heavy base64 image data while preserving code logic.

    Yes, the core profiling script is written in zero-dependency Python 3.9+. While it can use libraries like PyTorch or Polars for deeper analysis, the 'context shield' and header-parsing features work out-of-the-box on standard Python environments.

    You will receive the 'sci_eda_probe.py' core script, documentation for CLI integration, and structured JSON output templates for your agent's pipeline. All future updates to format support and parsing logic are included with your purchase.

    Unlike general EDA scripts, this is specifically engineered for LLM context windows with a hard 6 KiB output cap. It prevents 'hallucinated' data shapes by reading binary headers directly rather than relying on the agent's guesses or partial file reads.

    New in Research & Analysis