The Agent Skills Directory

Command Execution: The skill references an internal script, scripts/native-port.mjs, to perform its audit tasks. This is standard behavior for a tool designed to automate static analysis of a codebase.
Indirect Prompt Injection Surface: The skill is designed to analyze external data, which introduces a potential surface for indirect prompt injection.
Ingestion points: The agent processes various iOS project artifacts, including Info.plist, entitlements, and source code files within a declared target directory as specified in SKILL.md.
Boundary markers: The skill includes a detailed "Safety contract" and a "Required scope" section that establish clear boundaries for the agent's operations.
Capability inventory: The skill is restricted to read-only static analysis. It is explicitly instructed not to edit source code, create credentials, install new tools, or perform network uploads.
Sanitization: The instructions emphasize deterministic checks and evidence-backed reporting, which helps mitigate the risk of following instructions embedded within the files being audited.

audit-ios-readiness