The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements automated audits that scan project source files and built artifacts (e.g., main.jsbundle, Info.plist). While this logic processes external content, it uses strict regular expressions and defined rules rather than interpreting instructions. The scanner implementation includes built-in isolation to prevent data poisoning.
Ingestion points: The skill uses walkRepo and readFileSync in check-sim-setup.mjs and check-sim-app.mjs to ingest code and configuration files.
Boundary markers: The helper scripts/check-lib.mjs implements REPO_SCAN_IGNORES which automatically excludes agent skill folders, skill repositories, and generated build output from analysis.
Capability inventory: The skill utilizes execFileSync, spawnSync, writeFileSync, and rmSync to perform build and cleanup operations.
Sanitization: Input content is processed through maskComments and validated against specific patterns (e.g., UDIDs, bundle identifiers) before use.
[COMMAND_EXECUTION]: The build scripts and helper modules use execFileSync to execute standard macOS developer tools such as xcodebuild, xcrun simctl, and PlistBuddy. These operations are essential for the skill's functionality and are performed using validated arguments to ensure isolation on shared machines.
[DYNAMIC_EXECUTION]: The library script scripts/check-lib.mjs uses dynamic import() to load Node.js dependencies for specific tasks, such as processing PNG screenshots. The paths for these imports are resolved from a controlled tooling directory (.skill-tooling/) or environment variables.

ios-simulator-build