project-analysis

Trigger the analysis by asking to analyze the project or generate project documentation. It scans marker files like package.json, requirements.txt, pom.xml, Cargo.toml, and go.mod to detect languages, frameworks, build tools, and test frameworks.

The quality assessment scores six dimensions from 0 to 5: code organization, documentation completeness, testing, code style, security, and maintainability. Scores are averaged into an overall grade of A, A-, B, B-, C, or D with supporting evidence.

Yes, pass the focus_dir parameter to limit analysis to a specific subdirectory such as src/. Without it, the skill analyzes the entire project root while ignoring directories like node_modules, .git, dist, and build.

Projects are classified by file count: under 500 files uses deep mode, 500-2000 uses standard, and over 2000 uses quick mode. Quick mode retains core conclusions and risk items while reducing scan depth to stay within time and token limits.

Each run must produce one Markdown main report saved by default in the 文档 directory under the project root, named 项目分析-{timestamp}.md. You can override the location with the output_dir parameter.

The analysis degrades to a generic mode and still outputs the main report rather than failing. Dependency parsing failures are marked with their cause while other report sections continue normally.