skills/volcengine/searchcli/vs-search/Gen Agent Trust Hub

vs-search

Pass

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a packaged CLI tool (vs) to perform search operations and manage scene configurations. This is standard behavior for an AI agent skill interacting with a specific platform's infrastructure.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided search queries and configuration payloads which are interpolated into CLI commands. While this presents a surface for indirect prompt injection, it is considered a low risk in the context of a diagnostic and management tool.
  • Ingestion points: Search queries and scene configuration objects provided by the user or read from the environment (SKILL.md).
  • Boundary markers: None explicitly defined for preventing instruction injection within search queries.
  • Capability inventory: Execution of CLI commands including search run and search scene update (SKILL.md).
  • Sanitization: None explicitly mentioned; the agent relies on the underlying CLI tool's own validation and sanitization logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 24, 2026, 11:10 AM
Security Audit — agent-trust-hub — vs-search