The Agent Skills Directory

External Downloads (MEDIUM): The skill instructs users to install a package (omni-captions-skills) from a non-standard PyPI index hosted on GitHub Pages (https://lattifai.github.io/pypi/simple/). This source is not on the trusted repository list and may host unverified or malicious code.
Command Execution (LOW): The skill utilizes the Bash(omnicaptions:*) tool to run commands installed from the unverified source. While the skill includes a confirmation step for the user, the execution of external binaries from untrusted sources is a security risk.
Indirect Prompt Injection (LOW): The skill processes untrusted data from external video platforms which could contain malicious instructions.
Ingestion points: Video titles, descriptions, and captions from YouTube and 1000+ other platforms (SKILL.md).
Boundary markers: None present in the CLI usage; relies on manual user confirmation of the URL/Type/Path.
Capability inventory: Bash execution via omnicaptions, file read/write access via Bash and Read tools.
Sanitization: No evidence of sanitization or escaping for the downloaded metadata or captions before they are used in subsequent workflows (e.g., transcription or translation).

omnicaptions-download