external-collection

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the zilliz CLI tool to perform database operations including triggering, describing, and listing refresh jobs. These commands are restricted to the vendor-provided utility and follow standard operational patterns for the platform.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data returned from the service, such as job IDs and error messages, which are then presented to the user. This represents a standard feedback loop for asynchronous tasks.
  • Ingestion points: The skill reads the jobId from the trigger command and status/error strings from the describe command output in SKILL.md.
  • Boundary markers: None explicitly defined in the instructions, as the data originates from the trusted data-plane API.
  • Capability inventory: The skill's capabilities are limited to executing the zilliz CLI with predefined subcommands (external-collection refresh).
  • Sanitization: The skill instructs the agent to surface error messages verbatim to aid user troubleshooting of network or credential issues.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 06:55 AM
Security Audit — agent-trust-hub — external-collection