skills/g1joshi/agent-skills/trivy/Gen Agent Trust Hub

trivy

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSNO_CODE
Full Analysis
  • [COMMAND_EXECUTION] (HIGH): The skill provides documentation for executing the trivy command with user-controlled parameters. If an agent executes these commands using unsanitized user input (e.g., image names or local paths), it creates a vector for arbitrary command injection.\n- [EXTERNAL_DOWNLOADS] (MEDIUM): The skill demonstrates commands like trivy image and trivy repo which pull data from external, potentially untrusted sources. This allows external content to be introduced into the environment.\n- [INDIRECT_PROMPT_INJECTION] (HIGH): (Category 8) The skill's primary purpose is to process untrusted external data (container images, filesystems, and git repositories). This data is a known vector for embedding malicious instructions designed to subvert agent logic.\n
  • Ingestion points: trivy image [target], trivy fs [target], trivy repo [target] (SKILL.md).\n
  • Boundary markers: None present in the documentation.\n
  • Capability inventory: Shell command execution via trivy (SKILL.md).\n
  • Sanitization: Not addressed; the documentation encourages direct execution based on user-supplied targets.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 09:38 AM