The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several powerful system commands to manage the build environment.
Evidence includes process termination commands: killall -9 xcodebuild, killall -9 Simulator DeviceHub, and xcrun simctl shutdown all.
It performs recursive directory deletion to clear build artifacts and caches: rm -rf ~/Library/Developer/Xcode/DerivedData/*, rm -rf .build/, and rm -rf ~/Library/Caches/org.swift.swiftpm/.
These commands are legitimate within the context of troubleshooting Xcode environments and are restricted to standard development paths.
[EXTERNAL_DOWNLOADS]: The skill uses official tools to fetch external components from Apple's infrastructure.
Evidence includes commands like xcodebuild -downloadPlatform iOS and xcodebuild -downloadAllPlatforms to fetch simulator runtimes.
It also triggers package resolution via xcodebuild -resolvePackageDependencies, which downloads project dependencies from remote repositories specified in the project's configuration.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes untrusted data from the local environment.
Ingestion points: The agent reads project structures (.xcodeproj, .xcworkspace), build logs, and structured xcresult bundles.
Boundary markers: The instructions do not define explicit boundary markers for these inputs, though they recommend using structured tools like jq and xcresulttool to mitigate processing overhead and context flooding.
Capability inventory: The agent has the ability to execute shell commands, delete files, and terminate processes.
Sanitization: The skill uses jq to parse JSON output from simctl and xcresulttool to extract distilled error messages, providing a layer of structure over raw log parsing. This is considered a standard risk level for development-focused agents.

axiom-fix-build