asd-ste100-ru
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script (
scripts/utr_check.py) using thepython3interpreter. This script is used as a linter to verify that the rewritten text adheres to technical Russian standards (word count limits, sentence structure, etc.). The execution is performed locally on the agent's environment using standard parameters and heredocs to securely pass text input. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and processes user-provided text or files. However, it contains robust mitigations: the instructions explicitly command the agent to ignore any requests, commands, or formatting instructions found within the source text, treating them strictly as data. Additionally, the skill's workflow uses boundary markers (heredocs) when passing processed data to the validation script to prevent injection during the shell execution phase.
Audit Metadata