The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions allow for the processing of natural language strings in the query field of the get_rights task and the campaign.description field of the acquire_rights task.
Ingestion points: Request schemas for get_rights and acquire_rights in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided within the skill to isolate user-supplied natural language content from the agent's instructions.
Capability inventory: The skill is designed to retrieve sensitive data, including generation_credentials for voice synthesis and likeness, which could be leveraged if an injection is successful.
Sanitization: The documentation does not specify any validation or sanitization requirements for the free-text input fields.

adcp-brand