mauhpr/agentlint

Enforces rules across 8 context-aware packs (universal, quality, language/framework-specific, security, autopilot) that auto-activate based on project files or opt-in configuration. Integrates with Claude Code via hook events, blocking ERROR-severity violations and injecting contextual warnings into the agent's prompt to prevent drift like secrets commits, test skipping, and destructive commands. Designed for long-running agent sessions where traditional post-commit review arrives too late—catches infrastructure-level risks (unsafe shell execution, network exfiltration, force-pushes) in real time with millisecond-latency local rule evaluation.

14 stars and 1,341 monthly downloads. Available on PyPI.