stata

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides detailed instructions and examples for executing system-level shell commands using Stata's shell and ! syntax. This includes file system management, Git version control, and running the Stata binary in batch mode.
  • Evidence found in SKILL.md and references/external-tools-integration.md.
  • [REMOTE_CODE_EXECUTION]: Documents the execution of arbitrary Python and R code through the python: and rcall: environments. These features allow the agent to run scripts in separate runtime environments integrated with Stata.
  • Evidence found in references/external-tools-integration.md.
  • [EXTERNAL_DOWNLOADS]: Fetches Stata packages and datasets from official community registries (SSC) and established academic GitHub repositories.
  • Evidence: net install rdrobust, from("https://raw.githubusercontent.com/rdpackages/rdrobust/master/stata") in packages/rdrobust.md.
  • [DATA_EXFILTRATION]: Documents capabilities to connect to external SQL databases (ODBC/JDBC) and ingest data from remote URLs, which could be leveraged for data access or unauthorized export.
  • Evidence found in references/data-import-export.md and references/external-tools-integration.md.
  • [PROMPT_INJECTION]: The skill documents an indirect prompt injection surface by providing instructions for ingesting untrusted data from external files and databases without explicit sanitization guidelines.
  • Ingestion points: import delimited, import excel, and odbc load in references/data-import-export.md.
  • Boundary markers: Absent in documentation examples.
  • Capability inventory: Shell command execution, arbitrary code execution (Python/R), and file-write operations documented across multiple reference files.
  • Sanitization: No specific filtering or escaping mechanisms are discussed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 12:49 AM
Security Audit — agent-trust-hub — stata