The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources, such as creator bios, post captions, and video transcripts, through tools like social-fetch and watch-video. It proactively addresses the risk of indirect prompt injection by instructing the agent to treat this content as data for vetting rather than instructions to follow.
Ingestion points: External creator content retrieved during partner vetting (SKILL.md, social-fetch tool).
Boundary markers: Explicit instruction provided in SKILL.md to treat fetched content as untrusted data and ignore embedded instructions.
Capability inventory: Involves tools for fetching social media data and reading local agent context files (.agents/product-marketing.md).
Sanitization: Employs clear instructional guardrails to prevent the agent from executing content found in external transcripts or bios.
[SAFE]: The skill references a tool registry and integration guides hosted on a public GitHub repository (coreyhaines31/marketingskills). These links are for informational purposes and do not trigger automated downloads or execution of remote code.

influencer-marketing