The Agent Skills Directory

Command Execution: The skill executes a local Python script (build_explorer.py) and potentially other command-line tools like rsvg-convert or headless browsers (Playwright) to process data and generate visualization files. This is a core component of the skill's functionality.
Indirect Prompt Injection Surface: The skill is designed to ingest and process external threat model artifacts. This represents a potential surface where instructions embedded in the input data could attempt to influence the agent's behavior during the analysis or correctness review phases. The instructions include a manual verification process to mitigate this risk.
Local Network Binding: The skill instructions suggest binding a local HTTP server to 127.0.0.1 for previewing the generated HTML visualizations. This operation is restricted to the local loopback interface and is intended for quality assurance.
Data Ingestion and Sanitization: The build_explorer.py script reads user-supplied artifacts and includes logic to escape HTML special characters when embedding the content into the generated explorer template. This helps mitigate potential cross-site scripting (XSS) risks in the resulting standalone HTML file.
Mandatory Evidence Chain (Category 8):
Ingestion points: The skill reads external threat model artifacts (e.g., threat-model.md) through the build_explorer.py script and the agent's own reading of the source files in Phase 1.
Boundary markers: The skill relies on semantic IDs and explicit labeling of verbatim excerpts to manage data boundaries, though no strict technical delimiters are enforced on the raw input text.
Capability inventory: The skill utilizes file system access (read/write), local command execution via Python, and local network binding for previews.
Sanitization: The embedded function within scripts/build_explorer.py performs basic HTML character escaping before data is interpolated into the output HTML template.

threat-model-visualize