Loonghao Vx Project Analyze

Project Analyze Skill

This skill provides a systematic workflow for testing vx-project-analyzer against real-world projects.

When to Use

• Testing the project analyzer against specific open source projects
• Validating analyzer changes work correctly across different project types
• Discovering edge cases and missing features in the analyzer
• Iterating on fixes until analysis works correctly

Workflow

Step 1: Project Discovery

To find the project repository:

1. If the project name is ambiguous (e.g., "codex"), search GitHub to find the correct repository
2. Common project mappings are documented in references/known-projects.md
3. Verify the repository URL before proceeding

Step 2: Clone Project

To clone the project to a temporary location:

# Windows
cd c:/github && git clone --depth 1 <repo-url> <project-name>-test

# Unix
cd /tmp && git clone --depth 1 <repo-url> <project-name>-test

Use --depth 1 for shallow clone to save time and space.

Step 3: Run Analysis

To run the project analyzer:

cargo run --manifest-path c:/github/vx/Cargo.toml -p vx-project-analyzer --example analyze_project -- <project-path>

Step 4: Evaluate Results

To evaluate the analysis results, check:

1. Ecosystems detected - Are all languages in the project detected?
2. Dependencies - Are dependencies correctly parsed? Count should be reasonable.
3. Scripts - Are common scripts detected? No false positives?
4. Required tools - Are the right tools identified? No misidentified npm scripts?

Common issues to look for:

• Monorepo projects with language files in subdirectories
• Missing language support (e.g., Go, Java)
• False positive tool detection from internal script references
• Workspace/monorepo dependencies not being counted

Step 5: Fix Issues

If issues are found:

1. Identify the root cause in vx-project-analyzer code
2. Make targeted fixes to the relevant modules
3. Re-run analysis to verify the fix
4. Repeat until analysis is correct

Step 6: Cleanup

After testing is complete, remove the cloned project:

# Windows
Remove-Item -Recurse -Force c:/github/<project-name>-test

# Unix
rm -rf /tmp/<project-name>-test

Quick Reference

Supported Languages

Common Test Projects

See references/known-projects.md for a curated list of test projects.

Analysis Output Format

=== Analyzing: <path> ===

📦 Detected ecosystems: [<ecosystems>]

📋 Dependencies: <count> found
    - <name> (<version>) [<ecosystem>]
    ...

📜 Scripts: <count> found
    - <name>: `<command>`
    ...

🔧 Required tools:
    - <tool>: <description>
    ...

Troubleshooting

No ecosystems detected

• Check if marker files exist in root or immediate subdirectories
• Verify max_depth config allows subdirectory scanning

Dependencies count is 0

• For workspace projects, dependencies may be in member crates
• Check if the dependency parser handles the file format correctly

Too many required tools (false positives)

• Check if internal script references are being filtered
• Verify known_scripts context is passed to parser

Missing language support

• Check if a language analyzer exists in languages/ module
• May need to implement a new analyzer

Source: loonghao/vx — distributed by TomeVault.