The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves analyzing and providing feedback on user-supplied research grant components, such as Specific Aims pages and Project Summaries. This creates a potential surface for indirect prompt injection if the user-provided data contains instructions intended to override agent behavior.
Ingestion points: External text from research proposals is provided by the user to the agent for review and editing (e.g., 'Review my NIH Specific Aims page').
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing user-provided grant text.
Capability inventory: The skill employs the Read, Write, Edit, and Bash tools to assist with document management and project tracking.
Sanitization: There are no explicit measures described to sanitize or filter user-provided content before processing.

research-grants