The Qt Company has introduced a QML Profiler Skill for agentic development that lets AI agents handle performance profiling of Qt Quick 2D applications. The skill automates the full profiling workflow: locating Qt tools, enabling QML debugging, running the profiler in one of four modes (full, rendering, logic, memory), parsing the resulting .qtd trace file with a bundled Python parser, identifying hotspots mapped back to source code, and generating a timestamped Markdown report with frame-time percentiles, memory breakdowns, and specific fix suggestions. It works with Claude Code CLI, Claude Desktop, and GitHub Copilot in VS Code, and supports models like Claude Sonnet 4.6, GPT 5.4, and Gemini 3.1 Pro. Limitations include no support for Qt Quick 3D, GPU-side costs, C++ backend profiling, or hardware-specific optimizations. The skill is available on GitHub and Claude's marketplace under the 'qt-development' plugin.
Table of contents
When is the code profiling skill triggered?Step 1 - Locating the toolsStep 2 - Enabling QML debuggingStep 3 - Running the profilerStep 4 - Parsing the traceStep 5 - Analyzing hotspotsStep 6 - Writing the reportStep 7 — Console summaryLimitations of of the Code Profiling SkillDependenciesTested withGetting the skillSort: