The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes diagram content that may originate from untrusted sources or previous sessions without safety boundaries.
Ingestion points: The skill uses tools like get_diagram, search_diagrams, and list_diagrams to pull diagram data into the agent's context.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the diagram content as untrusted or to ignore any natural language instructions embedded within the Mermaid code.
Capability inventory: The agent has access to powerful tools including update_diagram, delete_diagram, and export_diagram (which notes the use of the Mermaid CLI), which could be targets of a successful injection attack.
Sanitization: No sanitization, validation, or filtering of the diagram content is mentioned before the agent reviews or modifies it.

continue-diagram