Reproducibility Quick Check
We scanned the repository's file structure, dependency files, and README against common reproducibility signals. This is a static surface-level check — for a deep, AI-driven code review, run a full audit.
Documentation
1/4README has usage instructions but no setup steps
No LICENSE file found
No CITATION.cff or citation file found
Dependencies
0/3No requirements.txt, pyproject.toml, or other pip/conda dependency file found
No lock file found and most dependencies lack version constraints
No .python-version or requires-python constraint found
Code Quality
1/4No .gitignore file found
No test files or tests/ directory found
No ruff, black, flake8, or other linter/formatter config found
Reproducibility
2/4No Dockerfile, docker-compose, or .devcontainer found
No mention of random seeds or reproducibility settings
The full audit goes deeper
Our AI agent explores your entire codebase to surface insights no static check can see.
Expert-level codebase exploration
Reads and reasons about your code the way a senior reviewer would — catches unclear docs, misleading comments, and subtle quality issues no static check can see
Context-aware, adaptive analysis
Every audit is custom-tailored — adapts its investigation based on what it finds, focusing on the files and patterns that matter for your specific project
Concrete fix suggestions for every issue
Provides specific, actionable recommendations for each finding so you know exactly what to change and why
Paper vs. code comparison
Verifies that your code implements what the manuscript claims
Clone-to-results walkthrough
Traces the full path from git clone to reproduced results
Hidden non-determinism detection
Surfaces subtle sources like set ordering, race conditions, and implicit sorts
Seed propagation tracing
Verifies all randomness sources are properly seeded, not just the obvious ones
External data link verification
Confirms that Zenodo, Figshare, and HuggingFace URLs actually resolve
Known vulnerability scan
Flags dependencies with published security advisories
Code complexity, dead code, and duplication detection
Identifies unmaintainable hotspots, unused code paths, and duplicated logic
Secrets and credentials scan
Catches leaked API keys, tokens, and hardcoded passwords
And more — with new checks shipping regularly
We're continuously adding new analyses based on the latest reproducibility research and user feedback
Unlock the full audit experience
A full audit takes about 10 minutes — it explores your codebase the way a senior reviewer would, tracing dependencies, verifying paper claims against code, and surfacing hidden issues with concrete fix suggestions.