flowchart-generator
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
Rendererclass inengine/renderer.pyis configured to downloaddraw.iodesktop installers directly from the officialjgraph/drawio-desktopGitHub repository. This behavior is intended to resolve compatibility issues with Visio exports by providing a specific supported version (v26.0.16). - [COMMAND_EXECUTION]: The skill uses
subprocess.runto manage the lifecycle of the flowchart rendering engine. This includes executing the downloaded installer for automated setup on Windows and macOS, as well as invoking thedraw.iocommand-line interface to perform image and document exports. Automated installation requires user confirmation via an interactive prompt before proceeding. - [PROMPT_INJECTION]: The skill is designed to ingest and process user-provided natural language business process descriptions, which represents a potential indirect prompt injection surface.
- Ingestion points: Step descriptions, roles, and document names extracted from user prompts and processed by the AI agent.
- Boundary markers: The skill documentation provides structured data schemas (JSON/Python dicts) to help the agent delineate untrusted content before passing it to the engine.
- Capability inventory: The skill has capabilities to write files (.drawio, themes), perform network downloads (official installers), and execute shell commands (rendering engine and installers).
- Sanitization: All user-provided text is sanitized using
html.escapeinengine/flowchart_builder.pybefore being embedded into the.drawioXML output, effectively mitigating injection risks within the generated files.
Audit Metadata