Coding Skill

Use this skill to generate and execute Python code for data processing, analysis, or computation tasks.

Most often the code result should produce an image of a plot or a textual output that answers the user's question. Those artifacts should be saved in scratch/

When to use this skill

• •Process or analyze scraped web content from scratch/ directories
• •Perform calculations or data transformations
• •Parse JSON/JSONL files from search results or URL content
• •Generate reports or visualizations
• •Any task requiring programmatic data processing

Available scraped data

The web skill saves content to the scratch directory:

• •scratch/query_*.jsonl - Search results (one JSON object per line)
• •scratch/url_*.jsonl - Scraped web page content with URL, title, and markdown content
• •scratch/USER_QUERY.txt - Original user question

Workflow

1. •Use generate_code to generate Python code, save it, and execute it in one step
2. •The code is automatically saved to scratch/code/ and executed
3. •Results are returned including output, exit code, and any new files created

Tools

• •grep_file(filepath: str, pattern: str, case_sensitive: bool, max_results: int) - Search for pattern in a file with regex
• •generate_code(task_description: str, context: str) - Generate Python code with Qwen, save it, execute it, and return results (all in one step)

Note: read_file, write_code, and run_code are disabled. Use grep_file to search files, or generate_code which handles everything automatically.

Tips

• •Always use relative paths starting with scratch/ to read data files
• •Scripts run with the project root as working directory
• •Use standard libraries (json, pathlib, etc.) without installation
• •For data analysis & viz, you can use pandas, numpy, plotly, matplotlib if needed
• •Print results to stdout - they will be captured and returned