The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): Analysis of scripts/graph_converter.py confirms it only performs local JSON-to-JSON transformations. It does not access sensitive system directories, environment variables, or hardcoded credentials, and lacks any networking code to exfiltrate data.
[Indirect Prompt Injection] (SAFE): The skill possesses a data ingestion surface via infranodus_output.json. 1. Ingestion point: infranodus_output.json read via json.load() in graph_converter.py. 2. Boundary markers: None. 3. Capability inventory: Only local file writing; no subprocess, exec, or network operations are present in any script. 4. Sanitization: The normalize_name function provides basic sanitization for entity IDs. The risk is negligible as the output is structured JSON and no execution follows.
[Dynamic Execution] (SAFE): Both graph_converter.py and workflow_analyzer.py use static logic. There is no use of eval(), exec(), pickle, or other dynamic loading mechanisms that would allow for runtime code injection or execution of untrusted input.
[Command Execution] (SAFE): The scripts do not use the os.system or subprocess modules. All operations are confined to internal Python logic and standard file I/O.

infranodus-orchestrator