The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing a local Python CLI module (python -m summarizer) to perform its tasks. This is the intended primary function and uses defined arguments for source, provider, and model configuration.
[EXTERNAL_DOWNLOADS]: The skill fetches content from external URLs including YouTube, Google Drive, and Dropbox to perform transcription. These operations are conducted through established libraries like pytubefix and youtube-transcript-api.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests data from untrusted external sources and interpolates it into LLM prompts.
Ingestion points: External video URLs (YouTube, Google Drive, Dropbox) and local video/audio files provided via the --source flag.
Boundary markers: Prompt templates in summarizer/prompts.json use the {text} placeholder for transcript insertion; however, the skill documentation does not specify the use of robust delimiters or 'ignore' instructions to prevent the model from obeying commands embedded within the transcript content.
Capability inventory: The CLI tool performs network requests to video platforms and LLM provider APIs (Gemini, Groq, OpenAI, etc.), and writes summary files to the local summaries/ directory.
Sanitization: No explicit sanitization, filtering, or validation of the extracted transcript text is mentioned before it is sent to the LLM for summarization.

Video Summarizer