Cliparr media editor/export workflow guidance for HLS playback, source selection, Mediabunny export, timeline normalization, subtitle burn-in, metadata tagging, media proxy behavior, and memory-sensitive export changes. Use when an agent changes or reviews files under apps/frontend/src/components/editor, apps/frontend/src/lib media/export helpers, apps/server media proxy/provider playback paths, or shared provider playback contracts.
Workflow for updating homepage marketing videos and static first-frame images. Includes ffmpeg and ImageMagick commands for extracting first frames and optimizing videos for web (MP4/H.264 and WebM/VP9) at specific resolutions for desktop and mobile.
Fully local free text-to-image skill for OpenClaw and general assets. Generates SVG from prompt, then converts SVG to PNG with local tools only (no online API calls).
Generate and edit images from the CLI using picture-it. Use this skill whenever the user asks to create, edit, or manipulate images — blog headers, social cards, hero images, product comparisons, YouTube thumbnails, movie posters, magazine covers, Instagram edits, background removal, or any visual content. Also trigger when the user mentions picture-it by name, wants to composite images, apply color grading, add text to images, remove or replace backgrounds, crop/resize photos, or needs any kind of image generation or photo editing from the terminal. This skill covers multi-pass AI image editing workflows that chain composable operations together.
Audit and improve CLAUDE.md files in repositories. Use when user asks to check, audit, update, improve, or fix CLAUDE.md files. Scans for all CLAUDE.md files, evaluates quality against templates, outputs quality report, then makes targeted updates. Also use when the user mentions "CLAUDE.md maintenance" or "project memory optimization".
Reverse engineer / decode CAN bus signals: find which CAN ID, byte and bit encode a real-world quantity (vehicle speed, engine RPM, state-of-charge, coolant temp, throttle, door lock, ...) and derive its start-bit, length, endianness, scale and offset into a DBC. A deterministic chain of Python scripts (survey -> correlate -> bitsearch -> build_dbc -> verify) replaces "a human watches the screen to spot correlations". Works three ways: (1) OFFLINE - decode a signal from an EXISTING recorded log the user already has (a CAN + OBD2 log, an MF4/CANedge log, or a webCAN CSV) using a separately decodable reference such as an OBD2 PID (engine RPM, vehicle speed, coolant temp) or a CSS Electronics GPS source (CANmod.gps, or a CANedge's internal GPS/IMU on CAN9 - bundled DBCs, e.g. GPS speed as a reference for proprietary vehicle speed) - NO hardware needed; (2) LIVE - capture from a CSS Electronics CANsub interface with a human-supplied reference; or (3) VISION - the user has a CAN log plus a VIDEO of a display (dashboard / gauge app / instrument) showing the true value, and a local open-source OCR script digitizes the on-screen number into the reference. Trigger this skill whenever the user wants to reverse engineer, decode, identify or "find the CAN ID/bits for" an unknown signal, or build a DBC from a log - INCLUDING phrasings like "reverse engineer Vehicle Speed from my CAN/OBD2 log file <name>.csv", "which CAN message carries RPM in this log", or "decode <signal> from my CAN log using this dashboard video as the reference". Targets plain, non-multiplexed CAN signals. Ad-hoc input logs the user refers to by filename usually sit in the working-directory root (or under TEMP/) — glob for the named file there before asking where it is.
微信公众号排版。Use when the user asks to turn an article, Markdown draft, longform post, essay, newsletter, review, or Chinese article into a WeChat Official Account-ready HTML layout with local preview, selectable Claude/OpenAI/Google-inspired styles, and a button to copy rich HTML for pasting into the WeChat editor.