wraith

An extremely fast AI code linter, written in Rust.

Catches what your AI forgot to check — hallucinated APIs, phantom packages, hardcoded secrets, AI artifacts, and supply chain risks. 20 rules, zero config.

$ wraith check .

src/api/handler.py:42:5  AG001 pandas.read_csv: unknown parameter "fast_mode"
                               → did you mean "mode"? (auto-fixable)
src/api/handler.py:78:1  PH001 import "ai-utils-secure": package not found on PyPI
                               → possible slopsquatting target
src/utils/auth.py:15:0   VC001 hardcoded secret: API_KEY = "sk-proj-..."
                               → known secret prefix; use os.environ["API_KEY"] instead (auto-fixable)
src/utils/auth.py:89:4   VC011 potential secret leak: 'api_key' passed to print()
                               → avoid logging or printing secret variable 'api_key'

Found 4 issues (2 auto-fixable). Run with --fix to apply.
Checked 128 files in 0.34s

Why wraith?

AI code generators hallucinate APIs, invent packages, and leave secrets in your code. Traditional linters don't catch any of it.

Scans 1121 files (FastAPI) in 350ms. Zero config, zero dependencies.

Install

pip install wraith

Quick start

wraith check .                       # scan current directory
wraith check . --fix --diff          # preview fixes as unified diff
wraith check . --fix                 # apply fixes
wraith rules                         # list all 20 rules

Features

• Hallucinated API detection — catches non-existent functions, kwargs, and deprecated calls via Python introspection
• Phantom package detection — validates imports against PyPI, catches typosquatting and slopsquatting
• Secret detection — layered analysis: known prefixes (sk-, ghp_, AKIA), Shannon entropy, bigram name classification, character class filtering
• Taint analysis — tracks data flow from os.environ to print()/logging sinks
• Supply chain checks — unpinned dependencies, missing lockfiles, dangerous files
• AI artifact cleanup — removes # Generated by Claude comments, debug code, pdb imports
• # noqa suppression — industry-standard inline suppression, Ruff/flake8 compatible
• SARIF output — --format sarif for GitHub Code Scanning and VS Code integration
• Confidence scoring — --min-confidence 0.8 to filter noise, every finding has a confidence score

Rules

AG — API Guard

PH — Phantom

VC — Vibe Check

Configuration

# Select specific rules
wraith check . --select AG,VC001

# Skip PyPI checks (offline)
wraith check . --offline

# Include test files (excluded by default)
wraith check . --include-tests

# Enable pedantic rules (VC003 print detection)
wraith check . --pedantic

# Only high-confidence findings
wraith check . --min-confidence 0.8

# CI/CD integration
wraith check . --strict --format sarif > wraith.sarif

Inline suppression

print("debug")           # noqa: VC003
API_KEY = "sk-secret"    # noqa: VC001
import pdb               # noqa

Python API

import wraith

results = wraith.check_source('API_KEY = "sk-secret"')
fixed = wraith.fix('import pdb\nbreakpoint()')

How it works

Research

Built on peer-reviewed research:

• AST Hallucination Guard (FORGE '26) — API validation via library introspection
• Package Hallucinations in LLMs — 20% of AI-generated imports are phantom packages
• Slopsquatting — supply chain attacks via hallucinated package names
• VibeGuard — AI code artifact hygiene (motivated by Claude Code source map leak)
• Secrets in Source Code (Saha et al. 2020) — character class distribution for secret detection
• Argus — hierarchical reference analysis for secret detection

Contributing

Found a false positive? Missing a rule? Open an issue — bug reports with code samples are the fastest way to improve wraith.

License

MIT