Skills
skills/duonglx/chanmayfoods/devops/Gen Agent Trust Hub

devops

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • PROMPT_INJECTION (HIGH): Vulnerable to Indirect Prompt Injection (Category 8) due to high-privilege capabilities combined with external data ingestion.
    • Ingestion points: Dockerfiles and cloud deployment configurations referenced throughout SKILL.md.
    • Boundary markers: None. There are no delimiters to isolate agent instructions from untrusted infrastructure code.
    • Capability inventory: Broad command execution access via gcloud, wrangler, and docker for modifying production environments.
    • Sanitization: None. Data from external configurations is used directly in shell commands.
  • REMOTE_CODE_EXECUTION (LOW): Uses curl | bash for software installation.
    • Evidence: curl https://sdk.cloud.google.com | bash in SKILL.md.
    • Note: Downgraded to LOW as the source (google.com) is a Trusted External Source.
  • EXTERNAL_DOWNLOADS (LOW): Performs global installation of the wrangler CLI.
    • Evidence: npm install -g wrangler in SKILL.md.
    • Note: Downgraded to LOW as it uses the trusted npmjs.com registry.
Recommendations
  • CRITICAL: Downloads and executes remote code from untrusted source(s): https://sdk.cloud.google.com - DO NOT USE
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 16, 2026, 10:54 AM