The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts that use the subprocess module to interface with the system xctrace utility for recording and exporting performance data.
scripts/record_trace.py executes xctrace record via subprocess.Popen.
scripts/instruments_parser/xctrace.py executes xctrace export and xctrace version via subprocess.run.
[INDIRECT_PROMPT_INJECTION]: The skill processes Xcode Instruments .trace files, which ingest external, potentially untrusted data into the agent's context.
Ingestion points: scripts/analyze_trace.py parses XML data exported from Instruments traces, which contains application-level logs and symbol names.
Boundary markers: Data extracted from traces is presented as structured JSON or markdown without explicit delimiters or instructions for the agent to ignore instructions embedded within the trace content.
Capability inventory: The skill can execute shell commands and perform file operations via its included Python scripts.
Sanitization: No sanitization is performed on user-controlled strings (like log messages or symbols) extracted from trace data.
[METADATA_POISONING]: The skill documentation and metadata contain significant amounts of fictional/misleading technical data presented as authoritative.
Evidence: Extensive references to non-existent OS versions (iOS 26, iOS 27), hypothetical design systems ('Liquid Glass'), and fictional hardware ('iPhone Duo').
Risk: This deceptive content can mislead the agent into providing incorrect technical advice or generating code for non-existent APIs, effectively poisoning the model's knowledge for SwiftUI development tasks.

swiftui-expert-skill