skills/exmergo/dex/transform/Gen Agent Trust Hub

transform

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to install the uv tool using a piped shell command from a well-known technology provider's domain.
  • Evidence in SKILL.md and scripts/run.py: curl -LsSf https://astral.sh/uv/install.sh | sh.
  • [EXTERNAL_DOWNLOADS]: The transformation engine is dynamically fetched from the vendor's package repository during the skill's execution.
  • Evidence in scripts/run.py: Use of uv run to install and run the exmergo-dex-core package.
  • [COMMAND_EXECUTION]: The wrapper script utilizes process replacement to transition execution to the transformation engine.
  • Evidence in scripts/run.py: os.execvp(cmd[0], cmd).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external dbt project files, SQL transformations, and semantic layer definitions.
  • Ingestion points: SKILL.md (via the --edits-file payload) and CLI arguments in scripts/run.py.
  • Boundary markers: Employs schema validation and dbt's own parser to verify changes before they are applied.
  • Capability inventory: Shell command execution through the uv wrapper script for dbt project management.
  • Sanitization: SQL models are restricted to single read-only SELECT statements, and literal credentials are explicitly refused.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 02:34 PM
Security Audit — agent-trust-hub — transform