AI SLOP Detector — VS Code Extension v2.8.0

Real-time AI-generated code quality analysis inside VS Code. Surfaces deficit scores, structural anti-patterns, ML signals, and actionable diagnostics without leaving your editor.

What's New in v2.8.0

• Python Advanced Patterns: god_function (>50 lines or complexity >10), dead_code (unreachable statements), deep_nesting (control-flow depth >4) appear as inline diagnostics with precise line numbers
• ML Secondary Signal: when a trained model is present, the summary diagnostic and status bar tooltip include slop probability and confidence
• Rebuilt scoring formulas: ICR uses density x complexity amplifier, status follows a single monotonic axis (CLEAN / SUSPICIOUS / INFLATED_SIGNAL / CRITICAL_DEFICIT), project LDR uses SR9 conservative aggregation

Features

Inline Diagnostics (Problems Panel)

Every analysis run produces structured diagnostics in the Problems panel:

Pattern diagnostics use pattern_id as the diagnostic code — VS Code's filter-by-code works natively (e.g., filter for god_function only).

Status Bar

Right side of the status bar shows live quality at a glance:

$(check) Good (12.4)      <- score below warn threshold
$(warning) Warning (34.1) <- score >= warnThreshold
$(error) Error (67.8)     <- score >= failThreshold
$(sync~spin) Analyzing... <- analysis in progress

Tooltip includes:

• Deficit Score and Status
• LDR Grade (A/B/C/D)
• LDR / Inflation / DDC metric values
• ML slop probability and confidence (when model is present)

Lint on Save / Lint on Type

• Lint on save (default: on) — triggers on every Ctrl+S
• Lint on type (default: off) — triggers with 1500ms debounce to avoid excessive analysis during active editing

Commands (Ctrl+Shift+P → "SLOP")

Installation

From VSIX (Local)

# Build
cd vscode-extension
npm install
npm run compile
npx vsce package --out vscode-slop-detector-2.8.0.vsix

# Install in VS Code
code --install-extension vscode-slop-detector-2.8.0.vsix

From Marketplace

Search "AI SLOP Detector" in the VS Code Extensions panel or:

ext install Flamehaven.vscode-slop-detector

Requirements

• Python 3.9+
• ai-slop-detector installed in the Python environment VS Code uses:

# Core only
pip install ai-slop-detector

# With JS/TS tree-sitter support
pip install "ai-slop-detector[js]"

# With ML secondary signal
pip install "ai-slop-detector[ml]"

# Everything
pip install "ai-slop-detector[full]"

The extension invokes python -m slop_detector.cli <file> --json internally. Set slopDetector.pythonPath if your Python is not on PATH.

Configuration

Open Settings (Ctrl+,) and search "SLOP Detector", or edit directly:

{
  // Enable / disable the extension
  "slopDetector.enable": true,

  // Trigger analysis on file save
  "slopDetector.lintOnSave": true,

  // Trigger analysis while typing (1500ms debounce)
  "slopDetector.lintOnType": false,

  // Show inline diagnostics
  "slopDetector.showInlineWarnings": true,

  // deficit_score >= failThreshold -> Error severity in Problems
  "slopDetector.failThreshold": 50.0,

  // deficit_score >= warnThreshold -> Warning severity in Problems
  "slopDetector.warnThreshold": 30.0,

  // Python interpreter to use
  "slopDetector.pythonPath": "python",

  // Optional path to .slopconfig.yaml
  "slopDetector.configPath": "",

  // Record results in local history database
  "slopDetector.recordHistory": true
}

Threshold Alignment with v2.8.0 Status Axis

The status axis in v2.8.0 uses:

Recommended threshold settings to match:

"slopDetector.warnThreshold": 30.0,
"slopDetector.failThreshold": 50.0

Diagnostic Reference

Pattern IDs (v2.8.0)

The following pattern_id values appear as diagnostic codes. Use VS Code's Problems panel filter to focus on specific patterns:

Structural:

• empty_function — function with only pass or ...
• bare_except — except: with no exception type
• god_function — function >50 logic lines or cyclomatic complexity >10
• dead_code — statements after return/raise/break/continue
• deep_nesting — control-flow depth >4

Placeholder:

• not_implemented — raise NotImplementedError
• mutable_default — mutable default argument (def f(x=[]))
• todo_comment — inline TODO/FIXME
• ellipsis_placeholder — ... as only function body

Cross-Language:

• js_var_usage — JavaScript var in Python file
• js_console_log — console.log in Python file
• ruby_array_each — Ruby-style .each block
• go_print_format — Go fmt.Printf style

ML Score Fields (when model present)

// Appears in summary diagnostic message and status bar tooltip
"ml_score": {
  "slop_probability": 0.82,   // [0,1] — probability of being slop
  "confidence": 0.91,          // [0,1] — model certainty
  "label": "slop",             // "slop" | "uncertain" | "clean"
  "model_type": "random_forest",
  "agreement": false           // true if ML and rule-based agree
}

When agreement is false, the rule-based deficit_score takes precedence.

Development

npm install
npm run compile     # One-time build
npm run watch       # Auto-recompile on changes

Press F5 in VS Code to open Extension Development Host with the extension loaded. The Output panel channel "SLOP Detector" logs all commands run and their stdout/stderr.

Changelog

See the main CHANGELOG for full history.

v2.8.0: Python Advanced patterns (god_function, dead_code, deep_nesting), ML score in diagnostics and tooltip, rebuilt ICR/status formulas.

v2.7.0: Docstring inflation diagnostics, evidence claim validation, hallucinated dependency detection, lint-on-type debounce.

License

MIT License — see LICENSE