django-verification

Pass

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a wide range of system and application commands (e.g., manage.py, pytest, ruff, bandit, pip-audit, git diff) to perform environment checks, security audits, and code quality assessments.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local project environment, including git diff output, project source code during linting, and security scan reports. This content is aggregated into a final summary report.
  • Ingestion points: Reads project files, git diff output, and tool reports (bandit-report.json, schema.json).
  • Boundary markers: The skill does not explicitly define delimiters to separate untrusted project content from agent instructions in the output template.
  • Capability inventory: The skill has capabilities to read files, execute shell commands, and generate network requests via security tools.
  • Sanitization: There is no evidence of sanitization or escaping for the content interpolated into the output report.
  • [DYNAMIC_EXECUTION]: Utilizes here-doc blocks (<< EOF) to pipe dynamically constructed Python scripts into the Django management shell for verifying database indexes, configuration settings, and logging functionality.
  • [COMMAND_EXECUTION]: Includes an example command for superuser creation that uses a hardcoded password ('admin'). While intended for test environments, this practice can lead to insecure deployments if copied verbatim to production scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 14, 2026, 03:02 PM