Schliff is an open-source linting, scoring, and verification harness developed by Zandereins designed for AI agent configuration ecosystems. It targets repository-level agent instruction files (such as AGENTS.md), agent skill meshes, and plugin specifications to ensure multi-agent instructions remain valid, non-colliding, and structurally sound. In day-to-day workflows, Schliff runs via a terminal CLI, pre-commit hooks, or CI/CD pipelines (such as GitHub Actions). Developers use it to inspect agent instruction health, check for overlapping skill definitions, and diagnose registry errors using subcommands like `schliff score AGENTS.md`, `schliff doctor`, and `schliff mesh`. For semantic evaluations, it also offers an optional LLM judge component backed by the Anthropic SDK. Schliff differentiates itself from runtime execution harnesses by acting as a dedicated static analysis and validation engine specifically tuned for agent architecture specs. Rather than executing prompts against a codebase, it validates the structural integrity and mesh definitions of agent toolsets and instructions before agents are run or merged. Schliff is open source and available as a Python package via PyPI, a pre-commit hook, and a GitHub Marketplace Action.
Tags: ai agent, cli, coding agent, developer tools, OpenSource, prompts