statechart
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONOBFUSCATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches the TLA+ model checker ('tla2tools.jar') from the official TLA+ project on GitHub and a Java runtime via the 'mise' tool manager (in 'scripts/run.ts'). The TLA+ JAR download is verified using a hardcoded SHA256 checksum, ensuring integrity.
- [COMMAND_EXECUTION]: Subprocesses are spawned for 'java', 'node', 'mise', and shell-based cleanup tasks (in 'scripts/statechart.ts' and 'scripts/run.ts'). Command arguments are constructed using sanitized numeric IDs and programmatically generated local directory paths, preventing shell injection.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external TLA+ specification and configuration data.
- Ingestion points: Specification ('.tla') and configuration ('.cfg') files provided as command-line arguments.
- Boundary markers: Absent; the content is parsed directly without delimiters or 'ignore' warnings.
- Capability inventory: Shell command execution via 'spawn' and 'execFile' to run the TLA+ engine.
- Sanitization: The skill implements custom-built tokenizers and parsers (e.g., 'scripts/frontends/config.ts') that validate input against a strict set of expected TLA+ symbols and tokens, preventing malicious payloads from influencing the execution environment.
- [OBFUSCATION]: Employs a custom encoding mechanism using specific character sequences (e.g., 'fl°°' and '¶ß' in 'scripts/backends/mermaid.ts') to handle character escaping for Mermaid and SVG output. This is a technical design choice to maintain character integrity during visualization and is not used for concealment of malicious logic.
Audit Metadata