vibesafe

Vibesafe is a specialized developer tool designed to bridge the gap between human intent and AI-generated code. It provides a robust contract system that allows developers to write Python function specs with type annotations and embedded doctests, which the AI then fills with verified implementations. By utilizing content-addressed hashing and deterministic checkpoints, Vibesafe ensures that the code running in production is exactly what was tested in development, preventing the 'drift' that often occurs with non-deterministic LLM outputs. This skill integrates the Vibesafe MCP server, allowing you to manage the entire toolchain—from scanning projects for decorated units to enforcing compliance gates and automating the regeneration of stale code.

• Performs comprehensive project scanning to identify and index all Vibesafe-decorated units and their current metadata.

• Automates code generation and recompilation tasks, supporting specific targets and force-regeneration for refined implementation control.

• Runs integrated doctests and quality gates to verify logic, type safety via mypy, and linting compliance via ruff.

• Manages and saves hash-locked checkpoints to freeze verified code, ensuring reproducibility across different environments.

• Generates detailed status reports covering versioning, unit counts, environment health, and checkpoint integrity.

• Facilitates drift detection by comparing current implementations against validated checkpoints, alerting developers to unauthorized spec changes before deployment.

• Use this skill when working on production-grade Python projects that require AI-assisted development without sacrificing safety or reproducibility.

• Inputs typically include target unit identifiers for selective operations or project-wide commands for bulk status checks and verification.

• Expected outputs include diagnostic tables, compilation logs, test pass/fail results, and confirmation of checkpoint synchronization.

• Keep in mind that Vibesafe enforces a strict dev/prod lifecycle: during development, mismatched hashes trigger auto-regeneration, whereas in production, these mismatches result in fast-fail errors to protect deployment integrity.

• Ensure your environment is configured with appropriate providers (e.g., OpenAI-compatible LLMs) to enable the automated generation capabilities within the Vibesafe framework.