The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python scripts to execute the xctrace system binary on macOS. This is used to list connected devices, record application performance traces, and export trace data for analysis.
Evidence found in scripts/record_trace.py and scripts/instruments_parser/xctrace.py using subprocess.run and subprocess.Popen with argument lists to mitigate injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes data from .trace files which contain application metadata such as view names and code symbols. There is a potential attack surface where a malicious application could include specially crafted view names or symbols designed to influence the AI agent's behavior when it parses the trace results.
Ingestion points: Performance data is ingested in scripts/analyze_trace.py after being exported from Instruments trace files.
Boundary markers: None identified in the script output; the agent processes raw view and symbol names.
Capability inventory: The skill can execute xctrace to launch or attach to processes and write files to the local disk (e.g., in scripts/record_trace.py).
Sanitization: The Python scripts parse XML data into structured JSON, but do not specifically sanitize string content for prompt injection patterns before delivery to the agent.

swiftui-expert-skill