write_file tool => no clear description on content

[nick-youngblut]

I've found that at least for Claude Haiku 4.5, the model will call the write_file tool many times in a row without providing the content parameter (3-8 times incorrectly before finally providing content). This issue might be alleviated by refining the tool description, currently in which content is never explicitly described as the content of the file:

WRITE_FILE_TOOL_DESCRIPTION = """Writes to a new file in the filesystem.

Usage:
- The file_path parameter must be an absolute path, not a relative path
- The content parameter must be a string
- The write_file tool will create the a new file.
- Prefer to edit existing files over creating new ones when possible."""

While the purpose of the content parameter might seem obvious, it could be interpreted by a naive LLM as "content type" parameter and thus not seen as a truly necessary parameter.

Also, the LLM might be reluctant to write out such a long parameter value to a tool; in my case, the content is >60k tokens. Anecdotally, the reluctancy to provide the content parameter seems to increase as the content size increases (e.g., 8-10 failed tool calls if the content token count is >100k tokens).

[quinlanjager]

I see this a lot too with Claude Sonnet 4.5. Seems to happen at a point where our models are producing a lot of content so this aligns with your observation.

[WenRichard]

I meet many times when I use claude-4.5

[marcofiocco]

I'm using deepagents==0.2.8 and can confirm that even Sonnet 4.5 calls write_file with just file_path, omitting content and therefore erroring:

1 validation error for write_file
content
  Field required [type=missing, input_value={'file_path': '/beam_defl...mgRaKGbKq', store=None)}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missingTraceback (most recent call last):


  File "/Users/fioccom/Documents/Code/JARVIS/jarvis-web/apps/jarvis-assistant/agent/.venv/lib/python3.12/site-packages/langchain_core/tools/base.py", line 891, in run
    tool_args, tool_kwargs = self._to_args_and_kwargs(
                             ^^^^^^^^^^^^^^^^^^^^^^^^^


  File "/Users/fioccom/Documents/Code/JARVIS/jarvis-web/apps/jarvis-assistant/agent/.venv/lib/python3.12/site-packages/langchain_core/tools/base.py", line 794, in _to_args_and_kwargs
    tool_input = self._parse_input(tool_input, tool_call_id)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


  File "/Users/fioccom/Documents/Code/JARVIS/jarvis-web/apps/jarvis-assistant/agent/.venv/lib/python3.12/site-packages/langchain_core/tools/base.py", line 672, in _parse_input
    result = input_args.model_validate(tool_input)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


  File "/Users/fioccom/Documents/Code/JARVIS/jarvis-web/apps/jarvis-assistant/agent/.venv/lib/python3.12/site-packages/pydantic/main.py", line 716, in model_validate
    return cls.__pydantic_validator__.validate_python(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


pydantic_core._pydantic_core.ValidationError: 1 validation error for write_file
content
  Field required [type=missing, input_value={'file_path': '/beam_defl...mgRaKGbKq', store=None)}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missing

[WenRichard]

I try the modify Prompt from merge request，but it still doesn't work，I guess Claude model cant cover long content because its output restrict，its curious in langchain agent！

[ollegreen]

+1 in deepagents==0.2.8

sometimes Sonnet 4.5 works after a couple of retries usually, but when running a longer deepagent with Opus 4.5: it can successfully write a new file in the beginning of the chain, but after a couple of sub-agent calls has close to 0% accuracy when trying to write a new file

breaks the use of deepagents in application development with these models

error from LangSmith:

1 validation error for write_file
content
  Field required [type=missing, input_value={'file_path': '/full_rese...ftfTXK86p', store=None)}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missingTraceback (most recent call last):


  File "/usr/local/lib/python3.13/site-packages/langchain_core/tools/base.py", line 891, in run
    tool_args, tool_kwargs = self._to_args_and_kwargs(
                             ~~~~~~~~~~~~~~~~~~~~~~~~^
        tool_input, tool_call_id
        ^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^


  File "/usr/local/lib/python3.13/site-packages/langchain_core/tools/base.py", line 794, in _to_args_and_kwargs
    tool_input = self._parse_input(tool_input, tool_call_id)


  File "/usr/local/lib/python3.13/site-packages/langchain_core/tools/base.py", line 672, in _parse_input
    result = input_args.model_validate(tool_input)


  File "/usr/local/lib/python3.13/site-packages/pydantic/main.py", line 716, in model_validate
    return cls.__pydantic_validator__.validate_python(
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
        obj,
        ^^^^
    ...<5 lines>...
        by_name=by_name,
        ^^^^^^^^^^^^^^^^
    )
    ^


pydantic_core._pydantic_core.ValidationError: 1 validation error for write_file
content
  Field required [type=missing, input_value={'file_path': '/full_rese...ftfTXK86p', store=None)}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missing

[williamchen433]

yes, does anyone have a solution for this ?

[nick-youngblut]

yes, does anyone have a solution for this ?

gpt-5.1 & gpt-5.1-mini

[WenRichard]

yes, does anyone have a solution for this ?

change model，like deepseek v3.1