review-areas
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONNO_CODE
Full Analysis
- Indirect Prompt Injection Surface: The skill aggregates information from external sources like pull request descriptions and commit messages to guide subagent reviews. This creates a surface where data provided by a user or external source is included in the subagent's instructions. * Ingestion points: Untrusted data enters the agent context in SKILL.md (Step 2) from PR descriptions and commit messages. * Boundary markers: The Area Prompt uses markdown headers to separate the change summary, though it does not explicitly instruct subagents to ignore potential instructions within that summary. * Capability inventory: The skill allows for reading files, running tests, and applying automated fixes to the filesystem as described in SKILL.md. * Sanitization: No specific sanitization or filtering of the external summary content is mentioned before its interpolation into subagent prompts.
- Workspace Command Execution: The skill instructions enable subagents to run tests and check diagnostics. This is a high-permission capability that allows the agent to verify code behavior directly. This behavior is expected in the context of a code review and testing utility.
- Automated Filesystem Modification: The 'review-and-fix' workflow allows the agent to investigate findings and apply changes to files. While this involves writing to the project, the skill includes a validation phase to verify that the applied fixes resolve identified issues.
Audit Metadata