auto-research
Fail
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes arbitrary shell commands defined in the research.toml configuration file via the run_command and setup_check parameters. The instruction to 'NEVER STOP' and 'NEVER ask permission to continue' removes human oversight from these operations.
- [REMOTE_CODE_EXECUTION]: The agent is instructed to implement code changes and then immediately execute them. This autonomous cycle of runtime code modification and execution allows for the execution of dynamically generated code without validation.
- [PROMPT_INJECTION]: The skill processes untrusted data from the project environment, such as command output in run.log, to guide its autonomous loop.
- Ingestion points: The agent reads run.log and project source files.
- Boundary markers: No delimiters or instructions to ignore instructions are used.
- Capability inventory: The skill can write to files and execute shell commands.
- Sanitization: No validation or sanitization of ingested content is performed.
Recommendations
- AI detected serious security threats
Audit Metadata