reviewing-devtools-prs
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external, untrusted content (pull request code and descriptions) which could contain malicious instructions.
- Ingestion points: Untrusted data enters the agent's context when it retrieves pull request content from the
flutter/devtoolsrepository (using GitHub CLI tools mentioned in the referenced parent skill). - Boundary markers: The skill explicitly refers to a 'strict user approval workflow' and human-in-the-loop review, which mitigates the risk of the agent performing unauthorized actions based on PR content.
- Capability inventory: The skill provides instructional guidance for code review (e.g., checking for disposal patterns, widget composition, and constant scoping). It does not include direct capabilities for file writes, network operations, or shell execution itself, relying instead on the agent's standard toolset.
- Sanitization: There are no explicit content sanitization or filtering steps described within the instruction set for handling external PR text.
Audit Metadata