build-entity-list

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes the tvly command-line tool to perform web research and uses python3 to handle the resulting JSON data. These operations are standard and aligned with the skill's purpose of aggregating web information.\n- [INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted external web data. It employs JSON schemas as a boundary mechanism to ensure that retrieved content conforms to a specific structure, which helps prevent malicious payloads from affecting agent behavior.\n
  • Ingestion points: Web research results obtained through tvly research.\n
  • Boundary markers: Use of the --output-schema flag to enforce data structure.\n
  • Capability inventory: Execution of search commands and a data-processing Python script.\n
  • Sanitization: Data is processed using JSON parsing to ensure structural integrity.\n- [DYNAMIC_EXECUTION]: A Python script is defined as a heredoc within a shell command to deduplicate entities found during search. This script is static and serves a specific, legitimate data-processing task without executing dynamically generated or external code.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 06:49 PM
Security Audit — agent-trust-hub — build-entity-list