retention-calculator
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (HIGH): The skill is designed to process external analytics content but lacks safety guardrails such as boundary markers or explicit instructions to ignore embedded commands.
- Ingestion points: Untrusted data files (SQL, CSV, logs) containing 'retention calculator' keywords.
- Boundary markers: Absent.
- Capability inventory: Bash, Write, Edit, Read, and Grep.
- Sanitization: Absent.
- [Command Execution] (HIGH): By explicitly enabling the 'Bash' tool, the skill provides a mechanism for remote code execution if the agent is influenced by malicious instructions embedded in the data it is tasked to analyze.
- [Data Exposure] (MEDIUM): The combined access to Bash and file-reading tools (Read, Grep) without restrictive pathing allows an attacker to read sensitive local files or environment variables via prompt injection.
Recommendations
- AI detected serious security threats
Audit Metadata