
from ponyflash-skill33
Generates images, video, speech, and music via the PonyFlash SDK and supports local media editing and subtitle workflows using FFmpeg.
PonyFlash Skill provides two capability families: cloud generation through the PonyFlash Python SDK (image, video, speech, music generation, model listing, file management, account/credits) and local media editing via an FFmpeg-based toolchain (clip, concat, transcode, extract audio, frame capture, subtitle prep). The skill guides API key setup and verification for cloud tasks, checks local FFmpeg capabilities for editing tasks, and contains playbooks for multi-step production workflows.
Use this skill when a user asks to generate AI media (images, video, TTS, music) with PonyFlash models or to perform local editing tasks—trimming, concatenation, transcoding, subtitle burning, and other FFmpeg operations. Choose the SDK path for cloud generation (requires API key) and the FFmpeg path for local edits. For end-to-end production, the skill coordinates both.
Useful with agents that can run Python and shell commands and that can upload/download files (Codex/Copilot-style agents or platform agents with Python runtime and ffmpeg).
PonyFlash is a dual-path media skill: cloud generation via the PonyFlash Python SDK (images, video, speech, music) and local media editing via FFmpeg scripts. 2 of 4 scripts ran successfully (ensure_subtitle_fonts.sh downloaded a font, media_ops.sh printed help); build_ass_subtitles.py needs required args and check_ffmpeg.sh fails because ffmpeg/ffprobe are not installed. The SKILL.md is well-organized with clear decision trees, security manifests in shell scripts, and atomic file writes.
ffmpegffprobeSecurity manifests in shell scripts are a nice touch. Atomic write pattern in media_ops.sh is well-implemented. Font download from CDN without checksum is a minor concern. No malicious patterns detected.