Skills
skills/dbt-labs/dbt-agent-skills/using-dbt-index/Gen Agent Trust Hub

using-dbt-index

Fail

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill contains instructions to install the dbt-index tool by downloading a shell script from the vendor's official domain (https://public.cdn.getdbt.com/fs/install/install-index.sh) and piping it directly to the shell (sh).
  • [COMMAND_EXECUTION]: The tool enables the execution of arbitrary SQL queries against local DuckDB indexes and remote data warehouses via the metadata run and warehouse run subcommands. It also includes system management commands such as system update and system uninstall.
  • [EXTERNAL_DOWNLOADS]: The skill downloads an installation script from the vendor's CDN and performs synchronization with dbt platform environments via the cloud-sync command.
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it processes dbt project artifacts.
  • Ingestion points: JSON artifacts (manifest.json, catalog.json, etc.) in the target/ directory.
  • Boundary markers: Explicit instructions are provided to treat tool output as untrusted and to avoid executing instructions found in metadata.
  • Capability inventory: The skill can execute shell commands (dbt-index), SQL queries, and network operations.
  • Sanitization: The skill relies on instruction-based boundaries rather than programmatic sanitization of metadata content.
Recommendations
  • HIGH: Downloads and executes remote code from: https://public.cdn.getdbt.com/fs/install/install-index.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
May 6, 2026, 07:18 AM