The Agent Skills Directory

[Filesystem Access Control]: The skill enforces a strict 'write boundary' that prohibits the agent from modifying files outside of an explicitly declared target directory. It specifically forbids writing to sensitive locations like the home directory or repository root, which helps prevent accidental or malicious file tampering.
[Bundled Guardrail Execution]: The instructions require running a bundled validation script (scripts/native-port.mjs authorize-write) to authorize every file modification. This serves as a localized security check to ensure changes are limited to approved scopes.
[Data and Credential Protection]: There are explicit prohibitions against copying sensitive artifacts such as credentials, private links, or customer data. The skill requires the use of synthetic fixture data and original or properly licensed assets, reducing the risk of data exposure.
[Environment Integrity]: The instructions forbid the agent from installing new software, runtimes, or accepting platform licenses. This constraint ensures the agent operates within its pre-defined environment and prevents unauthorized configuration changes.
[Source Code Processing Considerations]: As a skill that ingests React Native source code to generate new SwiftUI implementations, it possesses an inherent surface for indirect prompt injection. The skill mitigates this by requiring implementation in small, discrete units and strictly following a pre-defined parity contract rather than performing generic translations.

port-react-native-slice