statechart
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the tla2tools.jar package from the official Tlaplus GitHub repository (github.com/tlaplus/tlaplus) and installs the Java runtime (Java 25 via Eclipse Temurin) using the mise tool (jdx.dev) during its initial execution. These downloads originate from well-known sources associated with the skill's primary functionality and use checksum verification.
- [DYNAMIC_EXECUTION]: The script scripts/worker.ts dynamically imports and executes a module specified by the user via the --view command-line argument. This allows for the execution of code within that module to customize labeling and grouping during statechart generation.
- [COMMAND_EXECUTION]: The skill makes extensive use of subprocesses to run external tools, including the Java-based TLA+ model checker (TLC), the mise tool manager, and various JavaScript runtimes (Node.js, Bun, or Deno) to process model results and manage temporary build artifacts.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits surface area for indirect prompt injection as it ingests untrusted TLA+ specification files and TLC model checking output into the agent's context. While the skill primarily produces visualizations, the lack of explicit instruction boundary markers when processing these external inputs combined with its dynamic code execution capabilities represents a known risk vector.
Audit Metadata