thisis-romar

User-invocable front door for managing the OpenDeck Roadmap (GitHub project

Build and validate a reference .streamDeckProfile containing all known built-in Stream Deck action types for reverse engineering

Orchestrate the streamdeck-driver subagent to perform GUI operations in the Elgato Stream Deck app that cannot be done via JSON injection

End-to-end workflow for generating a .streamDeckProfile from an application's keyboard shortcuts

Fetch the spoken transcript (captions) of a YouTube video or Short. Use when the user gives a YouTube URL or video ID and wants the transcript, captions, spoken-content summary, or quotes from a video. Trigger with phrases like "get the transcript", "transcribe this YouTube video", or "what does this video say".

Clean up and structure a raw video transcript. Use when the user pastes or provides a messy YouTube/auto-caption transcript (with timestamps, rolling duplicate lines, VTT/SRT/json3 formatting, or broken mid-sentence wrapping) and wants it cleaned, de-duplicated, segmented into sentences or paragraphs, re-timestamped, or summarized. Trigger with phrases like "clean up this transcript", "fix these captions", "make this transcript readable", or "summarize this transcript". Runs offline with no network or extra packages.

Convert one image of a multi-view drawing sheet into per-view artifacts. Version 1.8.0 can read each view's region banner (US vs UK/EU) via --ocr-headers, extract drawing fields and right-side title blocks, emit clean drawings with the sheet-title band and view-label band stripped, auto-extract the gray legend/key box and itemize it into a structured bill of materials, vectorize clean drawings into scalable, layer-grouped SVGs, reconstruct complete sheets when a screenshot has cropped a title block (tagging every reconstructed field as inferred), warn about cropped title blocks on every run, and emit an authoritative per-run manifest with exact artifact counts and per-field provenance. Use for drawing views, plates, sheets, separated views, plan/elevation exports, downloadable per-view files, removing the title/header/footer, extracting the legend, or listing the fixtures/bill of materials from CAD/stage-layout sheets. Invoke on: "split this drawing", "separate the views", "one PDF per view", "give me the pla

Answer questions from a connected vault-brain MCP knowledge base (e.g. the grandMA2 manual) by querying the brain's tools instead of scanning files. Use whenever you need to find, look up, understand, or cite documentation, definitions, procedures, or reference data that lives in the vault. Encodes the tool-priority workflow: search → expand via graph → cite with breadcrumbs.