# lintlang

> Static linter for AI agent tool descriptions, system prompts, and configs. Catches vague instructions, missing constraints, schema mismatches, and other language-level failures before they cause agent misbehavior. Zero LLM calls. One dependency. Runs in CI.

lintlang uses the HERM v1.1 (Hermeneutical Evaluation and Risk Metrics) scoring engine to evaluate 6 dimensions of linguistic quality in agent configurations. On top of HERM scoring, 7 structural detectors (H1-H7) catch issues like empty tool descriptions, duplicate names, phantom schema fields, unbounded retry loops, and role sequencing errors.

Install with `pip install lintlang` (Python 3.10+). Use as CLI (`lintlang scan config.yaml`) or library (`from lintlang import scan_file`). Supports YAML, JSON, and plain text. Output as terminal, markdown, or JSON. CI integration via `--fail-under` flag.

## Documentation

- [README](https://github.com/roli-lpci/lintlang/blob/main/README.md): Install, usage, scoring, patterns, CI integration
- [Changelog](https://github.com/roli-lpci/lintlang/blob/main/CHANGELOG.md): Version history
- [AGENTS.md](https://github.com/roli-lpci/lintlang/blob/main/AGENTS.md): AI agent integration guide

## Optional

- [License](https://github.com/roli-lpci/lintlang/blob/main/LICENSE): Apache 2.0
- [PyPI](https://pypi.org/project/lintlang/): Package page
