The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/download_audio.py performs runtime installation or upgrade of the yt-dlp package using pip. While this is a common utility for media handling, automatic package installation at runtime is an unverifiable dependency pattern.\n- [COMMAND_EXECUTION]: Multiple scripts utilize subprocess.run to execute external tools and helper scripts. Specifically, download_audio.py executes yt-dlp and pip, while transcribe.py executes ffprobe for media metadata and json_to_text.py for data conversion.\n- [DATA_EXFILTRATION]: The skill transmits user-provided audio and video files to the TextOps API at text-ops-subs.com for processing. This is the core functionality of the skill, and the API key is transmitted via standard request headers.\n- [EXTERNAL_DOWNLOADS]: The skill performs several network-based downloads, including fetching version metadata from the vendor's GitHub repository in version_check.py and downloading media content from YouTube URLs in download_audio.py. It also allows uploading from user-provided URLs in transcribe.py.\n- [PROMPT_INJECTION]: The skill processes transcription results and YouTube titles which are sourced from untrusted third parties. This creates an indirect prompt injection surface. The skill instructions proactively address this by directing the agent to ignore any commands found within the transcribed text and to treat it strictly as data.

transcription-speech-to-text-hebrew