Hearing for the sandbox. Use for any task with audio reference files (.wav/.mp3/.flac/.ogg/.m4a/.aac/.aiff) or that asks to analyze, transcribe, measure, mix, master, or generate sound. Provides FFT / spectral-peak analysis, mel-spectrogram rendering, waveform sampling/decimation, integrated loudness (LUFS), tempo/BPM detection, resampling, and tone/clip synthesis.
Spreadsheet / tabular analysis and charting. Use for any task with tabular reference files (.csv/.xlsx/.parquet/.tsv/.json) or that asks to analyze, aggregate, model, forecast, or visualize data. Wraps pandas, numpy, matplotlib, and scikit-learn for loading, describing, charting, correlation, and quick linear regression.
Read and author office documents. Use for any task with document reference files (.pdf/.docx/.pptx/.xlsx/.csv) or that asks to produce reports, contracts, decks, spreadsheets, or memos. Extracts text and tables, renders PDF pages to images for visual inspection, and provides thin builders for DOCX/PDF/PPTX/XLSX deliverables.
Vision for still images. Use for any task with image reference files (.png/.jpg/.jpeg/.webp/.bmp/.tiff/.gif) or that asks to read text from an image (OCR), analyze colours, detect QR/barcodes, or transform pictures. Provides metadata, OCR, dominant-colour extraction, QR/barcode reading, and resize/grayscale/thumbnail helpers.
Vision for the sandbox. Use for any task with video reference files (.mp4/.mov/.avi/.mkv/.webm/.m4v) or that asks to watch, summarize, caption, storyboard, extract clips/frames, or analyze footage. Reads video frame-by-frame: metadata (fps/resolution/duration), uniform keyframe sampling, per-second frame extraction, scene-change detection, audio-track extraction, and contact-sheet montage rendering.