The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell-based tools including curl for API interaction, jq for JSON processing, and the vendor-specific CLI coli for audio transcription.
[EXTERNAL_DOWNLOADS]: Fetches user-provided media files and web content for processing. Downloads are restricted to temporary directories and occur only after input classification.
[DATA_EXFILTRATION]: Transmits user-provided content (URLs, text, and media) to the vendor's processing infrastructure at api.marswave.ai. This is documented as the primary function of the skill and does not involve access to sensitive system files or credentials.
[INDIRECT_PROMPT_INJECTION]: As a content processing tool, the skill possesses an attack surface for indirect prompt injection via user-provided materials and style references.
Ingestion points: Processes external URLs and user-supplied text during content extraction and style learning phases.
Boundary markers: Relies on the agent's internal instruction adherence; no explicit delimiters or 'ignore' instructions are visible in the high-level orchestration scripts.
Capability inventory: Includes file system writes, network requests via curl, and execution of the coli transcription tool.
Sanitization: No explicit sanitization of extracted material is performed before it is used to generate content.

creator