The Agent Skills Directory

[SAFE]: The skill uses defusedxml in scripts/instruments_parser/xml_utils.py for parsing XML data exported from Instruments. This is a security best practice that prevents XML External Entity (XXE) and entity expansion attacks when processing potentially untrusted trace exports.
[SAFE]: The analyze_trace.py script uses os.open with the os.O_EXCL flag and restrictive permissions (0o600) when creating report files. This ensures that analysis reports are created in a secure manner and are only readable by the user running the script.
[SAFE]: The record_trace.py script includes explicit validation for environment variable keys passed via the --env flag, ensuring they follow standard naming conventions and do not contain null bytes. Additionally, the script intentionally avoids printing the full command line to the console to prevent sensitive environment values (which might contain credentials) from leaking into logs.
[SAFE]: The SKILL.md instructions and scripts include clear security boundaries and warnings. For example, it advises the agent to treat .trace files as untrusted data and warns that launch environment values remain visible to process inspection, ensuring the agent and user are aware of the privacy implications of profiling.
[COMMAND_EXECUTION]: The skill's primary functionality involves executing the macOS xctrace utility through Python subprocess calls in scripts/record_trace.py and scripts/instruments_parser/xctrace.py. These calls are implemented using the list-based API (rather than shell strings) which effectively mitigates shell injection risks. The parameters (such as device UDIDs, process names, and app paths) are derived from user input but are passed directly to the binary as separate arguments.

swiftui-expert-skill