The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to create local work directories and initiate a Python HTTP server to host generated diagrams for the user. Evidence includes shell snippets in SKILL.md using mkdir and python -c "import http.server...". These commands are used for legitimate functionality and are scoped safely.
[INDIRECT_PROMPT_INJECTION]: The skill features an attack surface where untrusted user input is processed into an HTML visualization. Ingestion points: User-provided diagram descriptions used to generate SVG content and HTML titles in SKILL.md. Boundary markers: None provided in the instructions to separate user input. Capability inventory: File writing (write_file) and local server execution are used to materialize the output. Sanitization: The skill relies on the agent following the design system rather than explicit sanitization filters.
[SAFE]: The skill documentation provides explicit security guidance for local operations, requiring the use of '127.0.0.1' for the server bind address and recommending ephemeral ports to ensure isolation and prevent unauthorized access on shared networks.

concept-diagrams