The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides standard command-line instructions for iOS development triage, such as xcrun xctrace for profiling, leaks for memory analysis, and swift package resolve for dependency management. It also includes commands to clear local build caches (rm -rf ~/Library/Caches/org.swift.swiftpm) which are documented within the context of fixing Swift Package Manager resolution failures.
[REMOTE_CODE_EXECUTION]: While the skill discusses LLDB's ability to execute code at runtime using the expression command, this is presented as a fundamental debugging technique for modifying app state during a session. No patterns involving downloading and executing untrusted remote code were found.
[DATA_EXFILTRATION]: The skill mentions profiling network usage and inspecting HTTP payloads using the Instruments Network template. These are local diagnostic workflows for developers to optimize their own applications and do not involve unauthorized data transmission to external domains.
[PROMPT_INJECTION]: The instructions use standard instructional markers (e.g., 'IMPORTANT', 'CORRECT', 'DON'T') to guide the agent in providing debugging assistance. No attempts to override safety filters or bypass system instructions were identified.
[COMMAND_EXECUTION]: The skill provides guidance on configuring ~/.lldbinit to automate debugger setup, such as defining aliases and loading Python scripts for custom diagnostic commands (e.g., a view hierarchy dumper). These are standard power-user features of the LLDB debugger.

debugging-instruments