The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a potential surface for indirect prompt injection. Ingestion points: The instructions suggest fetching creator bios, posts, and comments for analysis via tools like social-fetch. Boundary markers: No specific delimiters are defined to isolate external data from the agent's instructions. Capability inventory: Capabilities are limited to analytical tasks and text generation; no high-privilege operations like file-writing or code execution are available to the skill. Sanitization: The skill relies on logical vetting processes for brand safety and authenticity rather than technical input sanitization.
[SAFE]: The skill demonstrates high security awareness by providing specific guidance on FTC disclosure compliance and identifying platform risks like inauthentic behavior. It includes evaluation cases that train the agent to refuse and correct non-compliant marketing tactics, such as undisclosed sponsorships.

influencer-marketing