The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied subtitle files (SRT/VTT) which can be a vector for indirect prompt injection if the files contain instructions targeting the AI agent.\n
Ingestion points: SKILL.md (Workflow Step 7) and scripts/check_captions.py ingest external subtitle data.\n
Boundary markers: The skill instructions do not specify any delimiters or warnings to ignore embedded instructions within the subtitle text.\n
Capability inventory: The skill utilizes local CLI execution for python and ffmpeg commands.\n
Sanitization: The check_captions.py script removes formatting tags via regex but does not filter or escape the actual text content for prompt injection patterns.\n- [COMMAND_EXECUTION]: The skill relies on executing a local Python script provided by the author and suggests standard ffmpeg commands for video processing tasks.

caption-styling