code-review-report

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several local commands as part of its defined workflow:
  • Uses git status and git diff to collect code changes for review.
  • Runs a bundled Python script using python3 <skill-dir>/scripts/inline_hljs.py <report-path> to embed a syntax highlighting library into the report.
  • Uses system commands (open, xdg-open, or start) to display the generated HTML file to the user.
  • [PROMPT_INJECTION]: The skill processes untrusted external data (Git diffs), creating a surface for indirect prompt injection.
  • Ingestion points: Reads code changes, commit messages, and untracked files via Git commands.
  • Boundary markers: Uses Markdown headers in subagent prompts to structure input, though it lacks explicit instructions to ignore embedded commands within the diffs themselves.
  • Capability inventory: The skill has access to the local filesystem and can execute Git, Python, and file-opening utilities.
  • Sanitization: Does not apply filtering or escaping to the diff content before processing it with subagents.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 11:59 AM
Security Audit — agent-trust-hub — code-review-report