The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from user-provided URLs and video metadata which could be leveraged to influence agent behavior through embedded instructions.
Ingestion points: External data retrieved and parsed via lib/url_parser.py (referenced in SKILL.md) and user-provided strings for titles, payoffs, and keywords.
Boundary markers: The instructions do not implement explicit delimiters or instructions to ignore or isolate potentially malicious content within the fetched data.
Capability inventory: The skill possesses a high-impact capability to publish generated content to an external platform through the lib.publish function call specified in SKILL.md.
Sanitization: There is no evidence of data sanitization, validation, or escaping logic applied to the content retrieved from external URLs before it is passed to the generation and publication steps.
[COMMAND_EXECUTION]: The skill interacts with local scripts and library functions to perform its core tasks.
Evidence: The skill invokes lib/url_parser.py to process external URLs and calls lib.publish to transmit data to the YouTube platform.

yt-description-writer