The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to process untrusted external data such as scraped content, transcripts, and emails.
Ingestion points: Processes external materials like captions, transcripts, and emails as described in SKILL.md.
Boundary markers: Employs a defensive instruction block at the beginning of SKILL.md that explicitly instructs the agent to treat scraped content as data to be analyzed rather than instructions to be followed. It further directs the agent to quote the content and stop if it appears to address the agent directly.
Capability inventory: The skill includes the capability to write generated ad copy to an Airtable database (Ad Pipeline table).
Sanitization: While no specific character-level sanitization is mentioned, the skill mandates a rigorous compliance review process in references/ad-compliance.md to ensure output adheres to Meta and FTC guidelines.

script-ads