The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data, including video topics and scripts, as well as external configuration from voice profile files. This information is used to generate captions and is passed to automated publishing functions.
Ingestion points: Processes user-supplied video scripts and topics, as well as data from ../../references/voice-profile.md.
Boundary markers: The skill does not employ explicit delimiters or boundary markers to distinguish instructions from ingested data.
Capability inventory: The skill can read video files from the filesystem via the video_path parameter and performs network operations (uploading, scheduling, and image generation) through the lib.publish and lib.illustrate functions.
Sanitization: No validation or sanitization of input data is performed before it is processed or transmitted to the publishing service.
[DATA_EXFILTRATION]: The skill transmits generated captions and video files to the Publora publishing service. It also facilitates network requests for AI-generated visuals via the lib.illustrate helper. These network operations are consistent with the skill's stated purpose of social media automation and publishing.
[COMMAND_EXECUTION]: The skill facilitates the handling and upload of local video files from paths provided by the user to the publishing infrastructure.

tt-caption-writer