alterlab-denario

Warn

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The Denario.get_results() method is described as executing agent-generated code to perform computations and visualizations (found in SKILL.md and references/research_pipeline.md). This core capability allows the AI agent to perform arbitrary operations on the host system depending on the methodology it develops.
  • [EXTERNAL_DOWNLOADS]: The installation instructions in references/installation.md reference a GitHub repository (AstroPilot-AI/Denario) and a Docker image (pablovd/denario) that do not match the author's stated identity ('AlterLab-IEU'). These are unverifiable third-party sources for the skill's core components.
  • [REMOTE_CODE_EXECUTION]: The skill facilitates the installation of a package (denario) that is designed to orchestrate agentic workflows with code execution capabilities. The installation process also involves fetching the Google Cloud SDK via a piped shell command (curl | bash), which is a standard procedure from a well-known service but contributes to the overall execution profile.
  • [PROMPT_INJECTION]: The skill processes untrusted text through set_data_description and set_method (found in SKILL.md). There is an absence of boundary markers or sanitization logic to prevent adversarial instructions in these inputs from manipulating the agent's reasoning or the code it generates for execution in the get_results phase.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 12, 2026, 12:47 AM
Security Audit — agent-trust-hub — alterlab-denario