The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill is designed to inspect and process external React Native or Expo feature files provided by the user.
Ingestion Points: The agent reads 'declared feature files' and 'directly imported files' from the repository (SKILL.md).
Boundary Markers: The skill includes a 'Safety contract' section that explicitly instructs the agent not to edit, create, or delete files, and to avoid copying private links or credentials.
Capability Inventory: The skill is scoped to inspection and documentation; however, it has access to read the file system and potentially invoke tools like the native-port.mjs script.
Sanitization: No specific sanitization of the source code is mentioned, but the output is intended to be a compact text-based behavioral contract.
Concern: Processing untrusted source code could expose the agent to embedded instructions (e.g., within code comments) that attempt to influence its analysis or output, though the skill's restrictive instructions mitigate the risk of unauthorized actions.
[Local Script Execution]: The skill instructs the agent to use a bundled local script (scripts/native-port.mjs) to validate files.
Context: The skill refers to this script as a 'guardrail' to check for ambiguity before proceeding with the planning phase.
Concern: The execution of a local script depends on the environment and the integrity of the files within the repository being analyzed. If the script itself is compromised or behaves unexpectedly, it could affect the agent's workflow.

plan-react-native-port