The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to automate performance profiling tasks. It interacts with Apple's xcrun simctl for simulator discovery and xcprof for recording and analyzing application traces. These actions are strictly scoped to the skill's primary function and include built-in safety boundaries like time limits.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external command outputs, such as application lists and performance analysis results. This surface is managed through clear instructions to use structured JSON output, reducing the risk of schema confusion or accidental instruction execution.
Ingestion points: Output from xcprof analyze --json, xcrun simctl list, and launchctl list commands.
Boundary markers: The agent is explicitly instructed to lead with user_frames and use JSON-parsing flags (--json) to ensure data is treated as structured information rather than instructions.
Capability inventory: Shell command execution (profiling tools, jq, grep, mktemp), file system access for trace storage, and application attachment.
Sanitization: The skill relies on structured data output from system tools to prevent command or prompt interpolation issues.
[SAFE]: The instructions establish a mandatory 'Consent Gate' requiring explicit user approval before the agent can launch arbitrary programs or record system-wide activity. Additionally, the skill provides guidelines for safe data containment using temporary directories and prevents destructive actions like automatic file deletion without confirmation.

axiom-profile-performance