The Agent Skills Directory

[Unverifiable Dependencies] (MEDIUM): The skill instructs users to install omni-captions-skills using an untrusted --extra-index-url (https://lattifai.github.io/pypi/simple/). This bypasses the default secure PyPI registry and could be used to distribute malicious code from an unverified GitHub Pages site.- [Indirect Prompt Injection] (HIGH): The skill processes external caption data and has the capability to execute commands via Bash, which is a high-risk combination for indirect prompt injection.
Ingestion points: Caption files (SRT, VTT, etc.) processed by the omnicaptions convert command.
Boundary markers: Absent; there are no instructions to prevent the agent from following commands embedded in the caption text.
Capability inventory: Access to Bash (restricted to omnicaptions:*) and ffprobe (via the --video flag).
Sanitization: No sanitization of caption content is implemented or described.

omnicaptions-convert