stata

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing and updating many third-party Stata packages from various community repositories (SSC, Stata Journal) and researcher-hosted GitHub pages (e.g., nppackages, rdpackages, sergiocorreia, haghish).
  • [COMMAND_EXECUTION]: Multiple reference files (e.g., references/external-tools-integration.md, references/workflow-best-practices.md) guide the user on executing system shell commands using the shell and ! prefixes, allowing for direct OS interaction and file system manipulation.
  • [REMOTE_CODE_EXECUTION]: The documentation includes extensive examples of Stata's bridges to external programming languages, specifically Python (via the python: command) and R (via the rcall package), which enable the execution of arbitrary code within those environments.
  • [DATA_EXFILTRATION]: The skill documents methods for retrieving data from external sources, including importing from remote URLs (import delimited "https://..."), interacting with web APIs using Python's requests library, and establishing connections to remote SQL databases via ODBC connection strings.
  • [INDIRECT_PROMPT_INJECTION]: The skill demonstrates how to ingest untrusted data from external URLs and databases into the agent's context. While standard for Stata, this creates a surface for indirect prompt injection if the ingested data contains malicious instructions, especially given the agent's documented ability to execute system commands and external scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 10:10 AM
Security Audit — agent-trust-hub — stata