The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides the agent with capabilities to read, search, and write academic documents, creating a vulnerability surface where malicious instructions embedded in user-supplied text could be interpreted as commands by the agent.
Ingestion points: The skill utilizes Read, Grep, and Glob tools to access and process the contents of external files within the workspace.
Boundary markers: The instructions lack specific delimiters or directive warnings to ensure the agent maintains context and ignores any adversarial instructions contained within the documents being edited or reviewed.
Capability inventory: The agent has permission to use the Write tool, which could be leveraged to modify local files if an indirect injection successfully overrides the agent's intent.
Sanitization: No mechanisms for sanitizing, escaping, or validating the content of ingested files are included in the skill's logic.

academic