apple-app-store-reviewer
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill interacts with local Apple developer tools (including simctl, xcodebuild, codesign, and otool) via the subprocess module. These calls are securely implemented using list-based arguments instead of shell strings, avoiding shell injection risks. Interactions are further protected by allow-lists for actions and regex validation of test identifiers.
- [EXTERNAL_DOWNLOADS]: User-initiated network requests are performed to fetch policy baseline updates from developer.apple.com and agentskills.io. This functionality is restricted to an allow-list of domains and requires an explicit --network opt-in from the user.
- [CREDENTIALS_UNSAFE]: A string matching the pattern for an OpenAI API key is present in tests/fixtures/project-bad/App.swift. This is a non-functional artifact within a test fixture used exclusively to verify the tool's ability to detect embedded secrets in analyzed user projects.
- [INDIRECT_PROMPT_INJECTION]: The skill audit process involves reading untrusted source code and metadata files which are then reviewed by an AI agent. This creates an indirect injection surface; however, the risk is mitigated by deterministic pre-processing and structured reporting.
- [DYNAMIC_EXECUTION]: The skill uses ast.compile() in scripts/validate_skill.py to perform syntax validation of Python scripts within the package. This is a standard validation practice and does not involve the execution of the code.
Audit Metadata