parallel-code-review
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes
bashto executegitandgh(GitHub CLI) commands. This is the primary mechanism for identifying code changes and retrieving PR metadata, which is essential for its core function as a code review tool.\n- [DYNAMIC_EXECUTION]: The skill executes a local Python scriptscripts/validate-review-output.pyto perform structural validation on reviewer feedback. It also uses theTasktool to dynamically spawn specialized agent reviewers and verification tasks.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data in the form of repository code diffs and PR metadata, which serves as the input for LLM-based reviewers.\n - Ingestion points: Source code files,
git diffoutput, andgh pr viewJSON data entering the agent context during Phase 1 and 2.\n - Boundary markers: The instructions do not specify the use of strict delimiters (e.g., XML tags) when passing code to reviewers, relying instead on the orchestration tool's default context management.\n
- Capability inventory: The skill has access to shell execution (
bash), file system reading (Read), and agent task delegation (Task).\n - Sanitization: While the skill implements a schema validation script for the output of the agents, it does not detail specific sanitization or escaping of the input code to prevent prompt injection attacks.
Audit Metadata