scripts
Warn
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPERSISTENCEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill allows for the execution of arbitrary shell commands by transmitting user-supplied command strings to a local daemon via the Unix domain socket specified in the
TARIBOY_TOOLS_SOCKETenvironment variable. - [PERSISTENCE]: Through its
schedulecommand, the skill provides a built-in mechanism for creating recurring tasks that run on a durable schedule, enabling the agent to maintain background activity across different sessions or iterations. - [DYNAMIC_EXECUTION]: The Python implementation in
scripts/scripts.pydynamically handles and assembles command-line arguments into instructions that are executed by the platform backend. - [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by ingesting and returning the output of shell commands and daemon responses to the agent context without sanitization or explicit boundary markers. Ingestion occurs in the
callfunction withinscripts/scripts.py, which reads JSON responses from the local socket; these responses are then displayed to the agent throughprint_result, potentially allowing untrusted command output to influence subsequent agent instructions.
Audit Metadata