The Agent Skills Directory

[SAFE]: The skill serves as a detailed knowledge base for performance marketing. It references standard industry tools such as Google Ads, Meta, LinkedIn, and TikTok, and provides legitimate educational resources. All external links lead to public GitHub repositories containing marketing tool registries and integration guides.
[INDIRECT_PROMPT_INJECTION]: The skill identifies that processing external data (such as competitor ads, landing pages, and customer reviews) creates an attack surface for indirect prompt injection. It includes specific mitigations in audit-guardrails.md and reading-google-ads-data.md, instructing the agent to treat all external inputs strictly as data and to never follow instructions embedded within them. These instructions serve as a defense against potential attacks.
[SAFE]: The static analysis hint regarding prompt injection is a false positive. The phrase 'ignore previous instructions' was detected within the security guardrails section, where it is used as an example of a directive the agent should ignore if encountered in untrusted data.

ads