The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes transcripts and metadata from YouTube videos, which are untrusted external sources. If a video contains malicious instructions within its subtitles, the agent could potentially be influenced when it processes that text.
Ingestion points: Video subtitle content is fetched via the redfox.hk API and parsed in the scripts/extract.py script.
Boundary markers: Absent; the script extracts raw transcript lines and presents them to the agent without specific delimiters or instructions to ignore embedded commands.
Capability inventory: The skill possesses the ability to read local configuration files (API keys) and write files to the user's ~/Downloads directory, including Markdown and Excel files.
Sanitization: Absent; there is no evidence of filtering or sanitizing the transcript text for prompt injection patterns before it is provided to the agent's context.
[COMMAND_EXECUTION]: The skill utilizes a local Python script, scripts/extract.py, to handle API communication, subtitle parsing, and file generation. This is a standard and expected execution model for this type of skill.
[EXTERNAL_DOWNLOADS]: The skill communicates with an external API at https://redfox.hk to perform its core function of subtitle extraction. This domain corresponds to the skill author's infrastructure (redfox-data) and is used for its intended purpose of providing video data processing. It also downloads subtitle files from YouTube's content delivery servers as specified in the API response.

youtube-digest