data-dictionary

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Stata .dta files, specifically extracting variable labels and dataset notes which could contain malicious instructions.
  • Ingestion points: External .dta files located via $ARGUMENTS (SKILL.md, Step 0).
  • Boundary markers: Absent; extracted metadata is interpolated directly into markdown templates.
  • Capability inventory: Executes Python scripts via subprocesses and writes files to the local file system (SKILL.md, Step 1 and 5).
  • Sanitization: No evidence of sanitization or escaping for the extracted metadata content before inclusion in the generated report.
  • [DYNAMIC_EXECUTION]: The skill generates and runs temporary Python scripts at runtime to perform data extraction using template-based logic (SKILL.md, Step 1).
  • [COMMAND_EXECUTION]: The skill uses shell-level operations for directory globbing to find data files and script execution via uv or python3.
  • [EXTERNAL_DOWNLOADS]: The skill requires the use of well-known Python libraries including pandas and pyreadstat from official package registries.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 08:52 AM
Security Audit — agent-trust-hub — data-dictionary