comprehensive-code-review
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell commands to analyze the repository state. It executes git commands (merge-base, rev-parse, diff, log) and file system utilities (find, grep, cat) to extract code patterns and framework failure semantics for the review process.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external sources which is then passed to sub-agents, creating a surface for potential instruction injection.
- Ingestion points: Pull request descriptions, commit messages, code diffs, and external requirements from Jira tickets and GitHub issues.
- Boundary markers: The orchestrator uses labeled placeholders like {DIFF} to structure the prompt, but it lacks explicit instructions for sub-agents to ignore potentially malicious directions embedded within the code or metadata.
- Capability inventory: The skill utilizes shell command execution and interacts with external APIs via the GitHub and Jira MCP servers.
- Sanitization: No validation or sanitization is performed on the content gathered from the git diff or external tickets before it is interpolated into the sub-agent prompts.
Audit Metadata