The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided research topics, questions, and draft papers, which creates an attack surface for instructions embedded within that data to influence agent behavior.
Ingestion points: External data is ingested through the $ARGUMENTS variable in SKILL.md, which can include full draft papers or research materials.
Boundary markers: The skill lacks explicit markers or instructions to isolate user-provided content from the agent's core instructions, potentially allowing the agent to follow directions hidden within research texts.
Capability inventory: The skill is configured with WebSearch, WebFetch, Write, and Task tools, which provide capabilities for network access and filesystem modification.
Sanitization: There is no evidence of input validation, escaping, or sanitization of the research materials before they are processed by the agent.

lit-review