Large File Write via Temp Script
Problem
Writing large files — especially TypeScript, TSX, or any content containing
Unicode emoji, backticks, ${} template literals, or multi-line heredoc
conflicts — can silently fail or truncate content when passed inline.
Failure modes observed:
write_filereturns[ERROR] unknown erroron large payloadsnode -e '...'with backtick template literals silently empties the target filepython3 -c '...'breaks on unescaped quotes or special charactersshell_agentmay time out or fail when the inline content is too large
Solution: Write a Python Script to a Temp File First
Instead of passing file content inline, write a small Python helper script to
/tmp/write_task.py (or similar), then execute it. The helper script holds
the target file content as a raw Python string or via base64 decoding.
Approach A — Python Triple-Quoted String (preferred for readability)
Use write_file or a heredoc to create the helper, then run it:
# Step 1: Write the helper script to /tmp
cat > /tmp/write_task.py << 'PYEOF'
import pathlib
content = r"""
// Your TypeScript / file content goes here
// Backticks `like this`, ${template} literals, and emoji 🎉 are all safe
export const MyComponent = () => {
const msg = `Hello, world!`;
return <div>{msg}</div>;
};
"""
pathlib.Path("/path/to/target/file.ts").write_text(content.lstrip("\n"), encoding="utf-8")
print("Written successfully.")
PYEOF
# Step 2: Execute the helper
python3 /tmp/write_task.py
Why
r"""? A raw triple-quoted string avoids all backslash escaping issues. The only content that cannot appear verbatim is"""itself — if needed, split the string or escape that sequence only.
Approach B — Base64-Encoded Content (most robust for binary/complex content)
When the content contains """ or other Python string edge cases:
# Step 1: Base64-encode your content (done once, outside the target system)
# python3 -c "import base64; print(base64.b64encode(open('file.ts').read().encode()).decode())"
# Step 2: Embed the base64 blob in the helper script
cat > /tmp/write_task.py << 'PYEOF'
import base64, pathlib
b64 = (
"aW1wb3J0IFJlYWN0IGZyb20gJ3JlYWN0JzsKCmV4cG9ydCBjb25zdCBNeUNvbXBvbmVudCA9"
"ICgpID0+IHsKICByZXR1cm4gPGRpdj5IZWxsbzwvZGl2PjsKfTsK"
# ... more lines if needed
)
content = base64.b64decode(b64).decode("utf-8")
pathlib.Path("/path/to/target/file.ts").write_text(content, encoding="utf-8")
print(f"Written {len(content)} bytes successfully.")
PYEOF
python3 /tmp/write_task.py
Approach C — shell_agent Delegation (simplest for agents)
Ask shell_agent to perform the write using the temp-script pattern:
"Write the following TypeScript content to
/src/components/MyPanel.ts. Do NOT usenode -eor inline Python. Instead write a Python script to/tmp/write_panel.pyusing a triple-quoted raw string, then execute it."
Decision Guide
| Situation | Recommended Approach |
|---|---|
Content has backticks / ${} template literals |
Approach A (raw r""") |
Content has """ sequences |
Approach B (base64) |
| Content is binary or non-UTF-8 | Approach B (base64) |
| Delegating to shell_agent | Approach C with explicit instructions |
| Quick small file | Direct write_file (no workaround needed) |
Anti-Patterns to Avoid
# ❌ DANGEROUS — backtick template literals will break node -e
node -e "const fs = require('fs'); fs.writeFileSync('out.ts', \`${content}\`);"
# ❌ FRAGILE — breaks on unescaped quotes and special characters
python3 -c "open('out.ts','w').write('${content}')"
# ❌ UNRELIABLE for large payloads — may silently truncate
echo "${large_content}" > out.ts
Verification Step
Always verify the write succeeded:
# Check file exists and has non-zero size
python3 -c "
import pathlib
p = pathlib.Path('/path/to/target/file.ts')
assert p.exists() and p.stat().st_size > 0, f'Write failed: {p}'
print(f'OK — {p.stat().st_size} bytes written to {p}')
"
Notes
- The temp script approach works for any file type:
.ts,.tsx,.js,.json,.md,.py, etc. - Always
lstrip("\n")the triple-quoted string to avoid a leading blank line. - For very large files (>50 KB), prefer base64 to avoid any risk of heredoc conflicts.
- Clean up temp scripts after use:
rm /tmp/write_task.py