arch-cross-service-integration
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION] (SAFE): The skill utilizes the Bash tool for troubleshooting tasks such as checking RabbitMQ queues and searching source code with grep. These operations are limited to local diagnostic commands consistent with the skill's stated purpose.
- [PROMPT_INJECTION] (LOW): Indirect Prompt Injection Surface: The skill instructions involve using grep to read source code and checking system logs via Bash. If an attacker can place malicious instructions inside the codebase or logs being analyzed, the agent could potentially be influenced.
- Ingestion points: Local source files (*.cs) and terminal output from rabbitmqctl.
- Boundary markers: Absent; no instructions provided to treat content in analyzed files as untrusted.
- Capability inventory: Bash, Read, Write, Edit, and Task tools are available to the agent.
- Sanitization: None observed; the agent directly processes the output of files and system tools.
Audit Metadata