The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The file scripts/render.ts includes an ensureDependency function that executes npm install beautiful-mermaid if the package is not found. This behavior downloads and installs code from a public registry at runtime without any version pinning or integrity checks.
[COMMAND_EXECUTION]: The skill utilizes child_process.execSync to run shell commands for dependency management, which could be exploited if the environment or package names were manipulated.
[EXTERNAL_DOWNLOADS]: The skill relies on npx and npm to fetch external code (tsx and beautiful-mermaid) from remote servers during execution.
[DYNAMIC_EXECUTION]: The rendering script employs dynamic ESM imports (import()) to load the beautiful-mermaid library after it has been potentially installed or updated at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by processing external data and possessing file system capabilities.
Ingestion points: Diagram source code is read from files or standard input in scripts/render.ts.
Boundary markers: There are no protective delimiters or instructions to prevent the agent from interpreting instructions that might be embedded within a processed Mermaid diagram.
Capability inventory: The skill can execute shell commands (execSync), read any file the user has access to (readFileSync), and write to arbitrary paths (writeFileSync) via the --output argument.
Sanitization: Input Mermaid code is passed to the rendering library without any filtering or sanitization steps.

mermaid