media-toolkit

Plugin media-toolkit from marketplace marcelerz in marcelerz/Claude. Commands read from the repo's manifest — 1 month ago.

Skills (12)

• build-manifest

  Use when the user wants to know how much of a content tree has actually been extracted — which PDFs still have no .txt, which audio has no transcript, what is stale — or wants an index of a processed corpus. Read-only coverage report over a tree the media skills have processed.

• check-media-tools

  Use when the user asks what media tools are installed, why a media skill reports a missing dependency, or how to install ffmpeg/yt-dlp/whisper/tesseract/pandoc for this toolkit. Probes every dependency the media skills need and reports what is missing, which skills need it, and the exact install com

• download-podcast

  Use when the user wants to download podcast episodes from an RSS/feed URL into a folder, or list what a feed contains. Fetches audio enclosures named "YYYY-MM-DD Title.mp3" and is safe to re-run.

• download-webpage

  Use when the user wants to save a web article/page as clean readable text or markdown (boilerplate stripped) into a folder. Uses trafilatura, falling back to curl+pandoc.

• download-youtube-videos

  Use when the user wants to download YouTube videos (a single URL, several URLs, or a list file) into a folder, optionally with subtitles or a whole playlist. Saves "Title [id].mp4" via yt-dlp and is safe to re-run.

• extract-all

  Use when the user wants to process an entire content tree at once — run the right extractor on every PDF, image, document, video, and audio file under a directory. Composite driver over extract-text + process-videos + transcribe-audio; safe to re-run.

• extract-audio

  Use when the user wants to extract the audio track from video files (a folder of videos or a specific subset) into mp3/m4a/wav/flac/opus. Uses ffmpeg and is safe to re-run.

• extract-text

  Use when the user wants to extract plain text from PDFs, scanned images, or documents (epub/docx/html/rtf) into sibling .txt files. PDFs use their text layer or adaptive OCR; images use OCR; docs use pandoc. Safe to re-run.

• extract-video-frames

  Use when the user wants still images out of a video — slides, on-screen code, schematics, title cards — or wants to OCR what is shown on screen rather than said aloud. Pulls frames with ffmpeg at an interval, a count, or on scene changes, and can OCR each one.

• process-videos

  Use when the user wants to fully process a folder of videos — extract audio, transcribe to SRT, and organize each video into its own named folder with a README. Composite pipeline over extract-audio + transcribe-audio; safe to re-run.

• summarize-transcripts

  Use when the user wants a summary, outline, chapter markers, or key moments from a transcript, subtitle file, or recorded talk. Normalizes .srt/.vtt/.txt into readable prose with a script, then writes the summary directly — the analysis is the model's work, not a tool's.

• transcribe-audio

  Use when the user wants to transcribe audio (mp3/wav/m4a — podcasts, interviews, extracted talk audio) to subtitles or text. Produces .srt (default), .txt, or .vtt with whisper.cpp; safe to re-run.

Similar plugins