autoresearch-create

Fail

Audited by Gen Agent Trust Hub on Apr 24, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill requires the agent to generate and execute bash scripts containing user-defined commands, enabling arbitrary code execution.
  • [PROMPT_INJECTION]: The instructions explicitly override agent behavioral constraints by mandating a 'LOOP FOREVER' mode and suppressing user confirmation prompts, stating 'Never ask should I continue?'.
  • [DATA_EXFILTRATION]: The skill allows the agent to read and modify files across the filesystem and includes a configuration option to set an absolute path as the working directory, which could lead to unauthorized data access when combined with autonomous execution.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 24, 2026, 06:32 AM