The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches transcription models from HuggingFace during its initial execution. HuggingFace is a well-known and established service for AI model hosting and distribution.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via python for transcription and ffmpeg for video encoding. This is the intended primary purpose of the skill for media processing tasks.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill incorporates user-provided file paths and 'theme names' when constructing output directories and shell command strings in SKILL.md.
Boundary markers: The command templates provided in the instructions do not explicitly include escaping or boundary markers for these user-controlled strings.
Capability inventory: The skill uses tools capable of executing shell commands and writing to the file system.
Sanitization: The instructions do not detail specific sanitization or validation logic for user-provided inputs (like the theme name) before they are interpolated into executable shell commands, which represents a potential injection surface if the executing agent does not apply its own safety controls.

auto-subtitle