streamflow-mypy-type-checking

name	StreamFlow Mypy Type Checking
description	This skill should be used when the user asks to "fix mypy errors", "add type annotations", "fix type checking", "resolve no-untyped-def", "fix mypy type errors", or when working with type hints while respecting StreamFlow's forbidden type constraints (no Any, dict[str, Any], etc.).
version	0.1.0

StreamFlow Mypy Type Checking Skill

Core Constraint

See AGENTS.md for the authoritative forbidden types list. If a fix requires a forbidden type, DO NOT FIX IT. Skip the error entirely.

Sub-Skills

Error code	Description	Sub-skill
[no-untyped-def]	Function missing type annotations	StreamFlow Mypy no-untyped-def Fixer skill
[var-annotated]	Variable missing type annotation	StreamFlow Mypy var-annotated Fixer skill

General Workflow

1. Classify: Can the fix use concrete types only? If NO → skip entirely
2. Fix: Add annotations (see sub-skill for patterns)
3. Validate — must return no output:

   git diff | grep -E "Any\b|MutableMapping\[.*Any|MutableSequence\[.*Any|list\[Any|dict\[.*Any"
   

4. Quality check — must pass:

   uv run make format-check codespell-check typing
   

5. Commit: Load the StreamFlow Git Workflow skill

Allowed Types Quick Reference

• Primitives: None, bool, int, str, float, bytes
• Concrete classes: Workflow, Connector, Port, Token, etc.
• Unions: str | None, int | bool
• Generics: list[str], dict[str, int], set[Token]
• Special: Self, type[MyClass], Literal["a", "b"]

Prefer Literal over str/int for fixed value sets (statuses, connector types, mode flags).

Broad arguments, narrow returns: parameters use abstract types (Iterable, Sequence, Mapping); return types use concrete types (list, dict, set).

Annotation Patterns

# Always annotate with concrete types
def process(self, workflow: Workflow, config: WorkflowConfig) -> bool: ...

# Self for classmethods
@classmethod
async def load(cls, persistent_id: int) -> Self: ...

# Optional: use X | None, not Optional[X]
def get(self, name: str) -> Connector | None: ...

# Always annotate return types — mypy strict mode requires it
async def close(self) -> None: ...

When in doubt, skip it. Better to leave an error than violate the forbidden types constraint.

See Also

• AGENTS.md — Forbidden types list, mandatory rules
• StreamFlow Git Workflow skill — Commit message format and approval workflow
• StreamFlow Code Style skill — Code style guidelines