The Agent Skills Directory

[DATA_EXFILTRATION]: The generate_uml and generate_diagram_url tools transmit diagram source code to the external service at kroki.io for rendering. While this is the intended functionality, users should be aware that architectural details, internal IP addresses, or network topologies included in the diagram code will be sent to this third-party domain by default.
[INDIRECT_PROMPT_INJECTION]: The skill ingests diagram source code via its primary tool parameters, creating a vulnerability surface where the agent might process and act upon malicious instructions embedded in data from external sources.
Ingestion points: The code parameter within the generate_uml and generate_diagram_url tools in SKILL.md.
Boundary markers: No delimiters or warnings instructing the agent to ignore instructions embedded in the diagram code were identified.
Capability inventory: The skill possesses file-writing capabilities on the local system through the output_dir parameter in the generate_uml tool (SKILL.md).
Sanitization: No evidence of input validation or sanitization for the diagram source code was found.

uml-diagram