FoundationAgents/MetaGPT · error · ValueError

{function_name} not found

Error message

{function_name} not found

What it means

SkillAction.find_and_call_function dispatches a skill by importing metagpt.learn and getattr-ing the function name on it. If the module import fails or the attribute does not exist (the skill was never registered/exported in metagpt/learn/__init__.py), it logs 'X not found' and re-raises as ValueError. Only skills explicitly exported by metagpt.learn (google_search, text_to_speech, text_to_image, text_to_embedding, skill_loader helpers) are callable.

Source

Thrown at metagpt/actions/skill_action.py:113

        try:
            rsp = await self.find_and_call_function(self.skill.name, args=self.args, **options)
            self.rsp = Message(content=rsp, role="assistant", cause_by=self)
        except Exception as e:
            logger.exception(f"{e}, traceback:{traceback.format_exc()}")
            self.rsp = Message(content=f"Error: {e}", role="assistant", cause_by=self)
        return self.rsp

    @staticmethod
    async def find_and_call_function(function_name, args, **kwargs) -> str:
        try:
            module = importlib.import_module("metagpt.learn")
            function = getattr(module, function_name)
            # Invoke function and return result
            result = await function(**args, **kwargs)
            return result
        except (ModuleNotFoundError, AttributeError):
            logger.error(f"{function_name} not found")
            raise ValueError(f"{function_name} not found")

View on GitHub (pinned to 11cdf466d0)

Solutions

1. Verify the function exists: run python -c "import metagpt.learn as m; print(hasattr(m, 'NAME'))".
2. Add your custom skill function to metagpt/learn/__init__.py (or monkeypatch/extend it) so getattr succeeds.
3. Fix spelling/case of skill.name to exactly match the exported python function name.

Example fix

# before
# my_skill defined only in my_pkg/skills.py -> ValueError: my_skill not found

# after
# metagpt/learn/__init__.py
from my_pkg.skills import my_skill  # now getattr(metagpt.learn, 'my_skill') works

import metagpt.learn as ml
assert hasattr(ml, skill.name), f"skill '{skill.name}' is not exported by metagpt.learn"

import metagpt.learn as _ml
def skill_exists(name: str) -> bool:
    return hasattr(_ml, name)

try:
    result = await SkillAction.find_and_call_function(name, args)
except ValueError as e:
    if 'not found' in str(e):
        logger.warning('skill missing, falling back')
        result = await fallback_handler(name, args)

When it happens

Trigger: SkillAction.run with skill.name set to a function not exported from metagpt.learn (e.g. 'my_custom_skill' or a misspelled name), or when a custom skill exists in a plugin package but was not added to metagpt.learn's namespace.

Common situations: Writing new skills in your own module but expecting SkillAction to find them; name mismatches between the skill yaml/registry name and the python function name; version changes that rename or remove a learned skill.

Related errors

• Trying to instantiate {class_full_name}, which has not yet b
• api_name: {api_name} not found
• Creator not registered for key: {key}
• Unknown config: `{type(key)}`, {key}

AI-assisted analysis of FoundationAgents/MetaGPT@11cdf466d0 (2026-08-14). Data as JSON: /api/errors/d0e0433e61d6edc5. Report an issue: GitHub.