code-reviewer

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts pr_analyzer.py, review_scope.py, and review_report_generator.py execute git commands and internal Python scripts via subprocess.run. The implementation uses list-based arguments rather than shell strings, which prevents shell injection attacks. This functionality is the primary mechanism for the skill's repository analysis and is used safely for local diagnostics.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of git diffs and source code files. The analysis results are compiled into reports that do not utilize specific boundary markers or delimiters for code snippets, creating a potential surface for indirect prompt injection where malicious instructions hidden in code comments could attempt to influence the agent's review verdict.
  • Ingestion points: Source code files and git diff outputs read by code_quality_checker.py and pr_analyzer.py during the review process.
  • Boundary markers: Absent. The final reports provide findings and code locations as plain text or standard markdown without explicit isolation markers.
  • Capability inventory: The skill's primary output is a structured review report and verdict (e.g., approve or block). These reports are intended to inform agent decisions or CI/CD gate logic, but the skill itself lacks network access or dangerous write permissions.
  • Sanitization: Absent. While the scripts scan for specific structural and security patterns using regex, the content is not sanitized against natural language instructions targeting the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 04:34 PM