The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/render_mermaid.py uses subprocess.run to invoke the mmdc (Mermaid CLI) tool. This is used to convert Mermaid diagram code into PNG images. The script executes the command with shell=False and builds the argument list programmatically, minimizing risks associated with shell injection.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several external dependencies, including matplotlib and numpy from the Python Package Index (PyPI), and @mermaid-js/mermaid-cli from the Node Package Manager (NPM) registry. These are well-established, trusted tools from official package repositories.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted legal analysis text from users to extract risk nodes and causal relationships (Step 1). While this creates an attack surface for indirect prompt injection, the agent serves as a structured parser, and the generated content is rendered into static images or Markdown reports, limiting the potential impact of adversarial content within the input data.

legal-risk-visualization