The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential indirect prompt injection surface as it is designed to ingest and process untrusted data from external website content, including footers, terms of service, and invoices, to determine target markets and operator locations. The skill provides logical boundary markers by instructing the agent to label evidence as OBSERVED, DECLARED, or INFERRED, which helps distinguish between different levels of data reliability. The agent's capabilities include writing to the .readyvibe/context.md file and browsing to official government and standards hubs. While the skill does not specify technical sanitization methods for ingested content, it contains strong instructional constraints that prohibit the agent from providing definitive legal advice or requirements from memory, effectively mitigating the risk of the agent being manipulated into making false safety claims.

jurisdiction-applicability