Skip to main content

Static Analysis

Goblin runs static analysis before the REVIEW stage to catch issues automatically.

Tools​

Ruff​

Fast Python linter:

  • Unused imports
  • Undefined variables
  • Style violations
  • Common errors

mypy​

Static type checker:

  • Type mismatches
  • Missing annotations
  • Incompatible assignments

Configuration​

# Enable static analysis (default: true)
goblin pipeline config -p PROJECT --config static_analysis_enabled=true

# Enable mypy (default: true)
goblin pipeline config -p PROJECT --config static_analysis_mypy=true

Execution​

Run automatically before spawning Reviewer:

# In pipeline_runner.py
static_analysis_context = self._run_static_analysis(pipeline)
prompt = static_analysis_context + "\n\n" + review_prompt

Output Format​

## Static Analysis Results

**Summary**: 2 error(s), 3 warning(s)

### Ruff Results

#### `app/models.py`
- Line 10: [F401] 'os' imported but unused
- Line 25: [E501] Line too long (120 > 100)

#### `app/views.py`
- Line 45: [W503] line break before binary operator

### mypy Results

#### `app/utils.py`
- Line 30: error: Incompatible types in assignment
- Line 55: error: Missing return statement

---

**Your focus areas** (tools cannot check these):
- Logic correctness and edge cases
- Breaking changes to existing functionality
- Security implications

Running Manually​

# Ruff
ruff check goblin/ --output-format json

# mypy
mypy goblin/ --json-output

# Combined
python -m goblin.core.static_analysis /path/to/worktree

API​

from goblin.core.static_analysis import (
    run_ruff,
    run_mypy,
    run_static_analysis,
    format_for_review_prompt,
)

# Run all
results = run_static_analysis(worktree_path)

# Format for prompt
context = format_for_review_prompt(results)

Result Structure​

@dataclass
class StaticAnalysisResult:
    ruff_findings: list[Finding]
    mypy_findings: list[Finding]
    total_errors: int
    total_warnings: int

@dataclass
class Finding:
    file: str
    line: int
    column: int
    code: str
    message: str
    severity: str  # error, warning

Reviewer Focus​

Static analysis handles mechanical checks, so Reviewer focuses on:

  • Logic correctness
  • Edge cases
  • Security implications
  • Breaking changes
  • Architectural concerns

Limitations​

Static analysis cannot detect:

  • Runtime behavior
  • Business logic errors
  • Integration issues
  • Performance problems
  • Security vulnerabilities (some)

The Reviewer's checklist covers these areas.