The Agent Skills Directory

[SAFE]: The skill includes an extensive 'Guard Stack' designed to sanitize and validate AI-generated output. This includes a 'truth judge' model and a deterministic grounding check that ensures every claim made by the AI is supported by the source advertisement text.- [SAFE]: Potential indirect prompt injection through ingested Facebook ad content is effectively mitigated by strictly scoping the AI's role to generating short, structured text and subjecting that output to multiple automated linting filters.- [SAFE]: Data collection is performed using well-known services (Apify and Google) and public website scraping. The skill enforces strict verification gates (root-page, title-match, and slug-match) to ensure it only processes data for the intended companies.- [SAFE]: The provided Python code snippets for platform integration are limited to logic, regex, and string processing. No dangerous execution patterns such as eval() or unauthorized network operations within the scripts were identified.

playbook-ad-library