js-run
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DYNAMIC_EXECUTION]: The skill uses the py-mini-racer library to evaluate Javascript code within a V8 isolate. This capability is intended for solving cryptographic challenges or performing complex calculations found in web pages. However, the execution of arbitrary, externally-sourced code is a significant capability that must be managed carefully by the agent.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of Javascript often retrieved from untrusted web environments. Because the results of these scripts are returned to the agent's context without sanitization, an attacker could craft a script that returns malicious instructions designed to manipulate the agent's behavior.
- Ingestion points: The source parameter in the run() method of the JsRun class and the environment parameter in the class constructor.
- Boundary markers: None present; the skill does not wrap inputs or outputs in protective delimiters to signal untrusted content.
- Capability inventory: The skill can execute any valid Javascript code within the memory-safe V8 isolate and return the result to the Python environment as native objects.
- Sanitization: No sanitization or filtering is performed on the Javascript output before it is passed back to the agent context.
- [EXTERNAL_DOWNLOADS]: The skill relies on the mini-racer Python package, which includes pre-compiled V8 binaries required for the JS execution environment.
Audit Metadata