Compile Context | Valstorm Docs

When the user provides a list of file paths and asks you to compile them into a context file (e.g., context.md), do NOT use the read_file tool to read their contents into your context window. This will cause massive token truncation and background delegation failures.

Instead, use a Python script executed via the write_file + terminal combination to read the files and write them directly to disk into a single Markdown file.

Trigger

The user provides multiple file paths.
The user asks to "compile context", "create a markdown file with these files", or "aggregate these files".

Steps

Identify the target output file path (e.g., vfs_context.md in the current directory or a specified path).
Identify the list of input file paths.
Write a Python script (e.g., generate_context.py) using write_file that reads each file and writes its content to the output Markdown file in a structured format.
Execute the Python script using the terminal tool: python3 generate_context.py.
Verify the file was created and inform the user.

Example Python Script Pattern

PYTHON

import sys

output_path = "/Users/jared/Documents/Code/monorepo/compiled_context.md"
files = [
    "/path/to/file1.py",
    "/path/to/file2.tsx"
]

with open(output_path, "w") as out:
    out.write("# Compiled Context\n\n")
    for f in files:
        try:
            with open(f, "r") as infile:
                content = infile.read()
            out.write(f"## {f}\n\n```\n{content}\n```\n\n")
        except Exception as e:
            out.write(f"## {f}\n\n```text\nError reading file: {e}\n```\n\n")

print(f"Context successfully aggregated into {output_path}")

Pitfalls

NEVER use read_file for this task. It will flood the agent's context window.
Ensure the python script handles file reading errors gracefully (e.g., file not found).
Use absolute paths in the Python script to avoid working directory confusion.