gs-statechart
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill automates the configuration of the TLA+ verification suite by downloading tools from established repositories. It fetches the TLC model checker from the official GitHub releases page of the tlaplus organization and uses the mise environment manager to handle the Java runtime. Artifact integrity is verified using SHA-256 checksums.\n- [DYNAMIC_EXECUTION]: The skill utilizes a modular design where the AI agent generates a JavaScript view script to format system states. The skill then imports this module and executes its functions to process model data. This involves the dynamic loading of executable content generated during the user session.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes console output from the model checker and interpolates it into bug reports and statechart labels. This design creates a surface where a malicious system description could lead to a model that prints instructions into the tool's logs. The generated bug reports currently lack structural delimiters or semantic filtering to isolate model output from the instructions interpreted by the agent.\n- [COMMAND_EXECUTION]: The skill orchestrates multiple components by spawning subprocesses for the TLA+ parser and model checker. It also includes a specialized janitor script to ensure robust lifecycle management, including process group termination and temporary directory cleanup.
Audit Metadata