job-agent
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
bin/rails consoleandbundle exec rspec. These tools allow the agent to execute arbitrary Ruby code and perform shell-based testing within the host environment, which are powerful capabilities. - [PROMPT_INJECTION]: The pattern for
ProcessImportJobinreferences/patterns.mddefines a workflow for ingesting external data. - Ingestion points: Data is processed from external sources via
import.file.downloadinreferences/patterns.md. - Boundary markers: There are no boundary markers or instructions to disregard embedded commands in the ingested data.
- Capability inventory: The skill can perform database writes, enqueue recursive background jobs, and send automated emails.
- Sanitization: No sanitization logic is included in the patterns to protect against prompt injection within the ingested CSV or JSON data.
Audit Metadata