The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data as part of its core functionality for vetting and analysis.
Ingestion points: The skill utilizes tools such as social-fetch and watch-video to ingest influencer bios, social media posts, and audience comments from external platforms.
Boundary markers: The instructions do not include specific delimiters to separate ingested data from agent instructions, relying on the analytical context.
Capability inventory: The skill can execute tools for data retrieval and read local context files like .agents/product-marketing.md.
Sanitization: No explicit sanitization of the retrieved external content is documented.
[SAFE]: The skill demonstrates security awareness by explicitly correcting non-compliant and risky marketing practices, such as undisclosed paid comments or hidden brand affiliations, in its reference playbooks.

influencer-marketing