The Agent Skills Directory

[PROMPT_INJECTION]: The skill design includes a surface for indirect prompt injection by allowing external data ingestion to influence agent behavior.
Ingestion points: External content enters the agent context via the sources parameter (which accepts URLs and text) in the create_playbook, create_entity, update_entity, and create_resource tools defined in SKILL.md.
Boundary markers: SKILL.md does not provide instructions for the agent to use delimiters or instructions to ignore embedded commands when processing content from external sources.
Capability inventory: The agent can perform significant modifications to the environment, including create_entity, update_playbook, and delete_resource, as documented in SKILL.md.
Sanitization: The skill instructions do not specify any validation or sanitization requirements for data fetched from remote URLs before it is processed by the model.

library