agentic-dev-team: Agent for Claude Code

Python Quality

Output JSON:

{"status": "pass|warn|fail|skip", "issues": [{"severity": "error|warning|suggestion", "confidence": "high|medium|none", "file": "", "line": 0, "message": "", "suggestedFix": ""}], "summary": ""}

Status: pass=clean Python, warn=improvements needed, fail=unsafe patterns Severity: error=bare except or type safety issue, warning=missing types or anti-pattern, suggestion=modern idiom Confidence: high=mechanical (add type hint, use f-string); medium=design choice; none=domain context needed

Context needs: diff-only File scope: *.py

Activates when

pyproject.toml, requirements.txt, or setup.py exists.

Skip

Return skip when no .py files in the changeset.

Detect

Type hints:

• Missing type annotations on function signatures (public APIs must have types)
• Any used without justification
• No mypy or pyright config for strict checking
• # type: ignore without explanation

Exception handling:

• Bare except: or except Exception: without re-raise or specific handling
• Silencing exceptions with pass
• Catching too broad (Exception when only ValueError is expected)
• Missing from in raise ... from chains

Modern idioms:

• format() or % string formatting instead of f-strings
• Manual dict/list construction instead of comprehensions
• type() checks instead of isinstance()
• Mutable default arguments (def f(x=[]))

Data modeling:

• Plain dicts where dataclass or Pydantic.BaseModel would add type safety
• Duplicate field definitions across multiple dicts
• Missing validation at API boundaries (use Pydantic for request/response)

Ignore

Django/Flask-specific patterns, test fixtures, script-only files, architecture.