Skills catalog

Portable Agent Skills with deterministic scripts, JSON contracts, and recipes. Filter by category or search with /.

14Skills
97Operations
Setup4 ops

install-media-tools

Detect, verify, and install media processing binaries (ffmpeg, ImageMagick, exiftool, tesseract, yt-dlp, uv) on macOS and Linux. Use before other mediaskills when doctor reports missing tools, when ffprobe/ffmpeg is not found, or when setting up a fresh dev machine or CI runner.

4 recipes
Inspect8 ops

inspect

Probe, describe, compare, and batch-inspect video, audio, and image files using ffprobe. Use when you need metadata (duration, resolution, codecs, size), want to compare two files before/after processing, or need structured probe data to answer questions about a media file without modifying it.

5 recipes
Audio9 ops

audio

Extract, convert, trim, concatenate, normalize, fade, resample, and silence-detect audio using ffmpeg. Use when you need to modify audio tracks, change format or sample rate, edit timeline segments, or prepare audio for muxing back onto video.

5 recipes
Image9 ops

image

Convert, resize, crop, rotate, flip, optimize, read EXIF, strip metadata, and OCR images using ImageMagick, exiftool, and Tesseract. Use when you need to transform still images, reduce file size, inspect camera metadata, remove sensitive EXIF before sharing, or extract text from screenshots and scans.

9 recipes
Video12 ops

program-master

TV program segmentation — detect fade-aware black+silent break separators, preserve frame-accurate fade anchors, label content segments, and generate validated thumbnail-led broadcast segment PDF reports. Use when finding commercial breaks, naming episodes/parts, mapping broadcast master structure, or delivering a visual segment report — not hard cuts inside content.

Video2 ops

shots

Detect hard cuts (shot boundaries) in video via ffmpeg scene detection and extract midpoint stills per shot. Use when the unit of analysis is a shot — cataloging cuts, QC, or feeding shot-level frames to vision/OCR — not TV program structure or commercial breaks.

3 recipes
Video11 ops

video-transformation

Transform video with ffmpeg — trim, concatenate, transcode, scale, proxy, extract frames and audio, create GIFs, replace audio tracks, and other stream manipulation. Use when changing video files through editing cuts, codec conversion, resizing, muxing, or derivative exports.

6 recipes
Acquire1 ops

download

Download video and audio from URLs using yt-dlp. Use when you need to fetch media from YouTube, Vimeo, or other supported sites before local processing with inspect, audio, or video-transformation skills.

3 recipes
Timecode9 ops

timecode

SMPTE timecode arithmetic with drop-frame and non-drop-frame support — convert between timecode and seconds, add/subtract offsets, retime between frame rates, convert DF↔NDF, and infer format from strings or ffprobe/MediaInfo metadata. Use when edit decisions are expressed as HH:MM:SS:FF or HH:MM:SS;FF or when aligning clips at NTSC rates.

9 recipes
Captions8 ops

captions-compliance

Broadcast caption compliance — validate and format SRT, export CEA-608 SCC and SMPTE-TT, flag busy-zone collisions, and transcribe speech to captions. Use when delivering FCC-aligned captions or handoff to linear/broadcast workflows.

7 recipes
Captions4 ops

speech-captions

Transcribe speech to timed captions with faster-whisper, detect spoken language, and build SRT/VTT from text or segments. Use when you need automatic captions from audio/video, language detection before transcription, or subtitle files from transcript JSON.

5 recipes
Captions4 ops

subtitles

Convert, shift, extract, and burn subtitle files (SRT/WebVTT). Use when you already have caption text or embedded subtitle tracks and need format conversion, sync adjustment, extraction from containers, or hardcoded captions on video.

5 recipes
Vision6 ops

forced-narrative-exact

Generate exhaustive, literal, frame-accurate forced-narrative reports from video masters containing burned-in dialogue subtitles. Use when a user requests a forced-narrative inventory, burned-in subtitle report, exact on-screen dialogue transcription, or start/end SMPTE timecodes for every subtitle, especially when embedded tmcd timecode, multiple program passes, textless duplicates, and exclusive frame boundaries must be handled consistently.

Vision14 ops

vision-analysis

Interval-based and shot-level visual analysis — extract frame sequences, guide agent vision analysis of stills, classified on-screen text, and QC-style reports. Use for burned-in subtitles, graphics, title cards, and timecoded text inventories. Does not bundle a vision model; the agent analyzes images with its own capabilities.

2 recipes
esc

Try trim, captions, timecode, or broadcast