The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection as it is instructed to fetch and process untrusted data from external sources and local documents.
Ingestion points: The agent uses WebSearch, WebFetch, and the Read tool to gather trademark research and regulatory guidance from the internet and local files.
Boundary markers: There are no instructions or delimiters provided to the agent to help it distinguish between its own system instructions and potentially malicious instructions embedded in the external data it reads.
Capability inventory: The agent has access to several powerful tools including Bash, Write, WebSearch, and WebFetch which could be manipulated if an indirect injection occurs.
Sanitization: The instructions do not specify any validation or sanitization steps for the data fetched via web tools before the agent processes it to draft documents.

scope-trademark