The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses macOS CLI tools for the full application lifecycle, including xcodebuild for compilation, xcrun for developer utilities, lldb for debugging, and instruments/xctrace for profiling. It also includes commands to modify system-level settings like the active Xcode developer directory via sudo xcode-select and network proxy settings via networksetup. Additionally, it suggests persisting convenience aliases by modifying shell configuration files like ~/.zshrc or ~/.bashrc (referenced in references/cli-workflow.md).- [EXTERNAL_DOWNLOADS]: The reference guides recommend installing several external development tools through package managers like Homebrew and RubyGems. These include xcodegen, mitmproxy, xcbeautify, xcpretty, and xcsift. It also integrates well-known libraries such as Alamofire for networking and Sparkle for application updates via Swift Package Manager (referenced in references/macos-polish.md and references/networking.md).- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes data from external sources that could be influenced by an attacker. Specifically, build logs, crash reports, and network traffic captured by mitmproxy are analyzed by the agent to diagnose issues.
Ingestion points: Build error logs from xcodebuild (processed via xcsift as described in references/cli-observability.md), runtime logs from log stream, crash reports in .ips format (referenced in workflows/debug-app.md), and network traffic dumps from mitmdump.
Boundary markers: Absent; the workflows do not specify clear delimiters or instructions for the agent to ignore potentially malicious content embedded within these tool outputs.
Capability inventory: Across its various scripts and reference files (e.g., references/system-apis.md, references/cli-workflow.md), the agent has the authority to execute shell commands, write to the filesystem, and perform network requests.
Sanitization: Absent; there is no evidence of sanitization or filtering applied to external content before it is processed by the agent.

build-macos-apps