vertex-ai-pipeline-creator

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • COMMAND_EXECUTION (HIGH): The skill metadata explicitly allows 'Bash(gcloud:*)' which permits arbitrary Google Cloud CLI commands. This is a high-privilege capability that could be abused if the agent is manipulated into executing commands outside the user's intent.
  • PROMPT_INJECTION (HIGH): The skill lacks instructions for sanitizing user input or defining clear boundaries for processed data. An attacker could provide a malicious pipeline definition or GCP configuration that, when processed by this skill, leads the agent to execute unauthorized gcloud commands (Category 8: Indirect Prompt Injection). Evidence: 1. Ingestion: User requests and pipeline configurations (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Bash(gcloud:*), Write, Edit (SKILL.md). 4. Sanitization: Absent.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 04:15 AM