lintlang

lintlang is a zero-LLM static linter for AI agent configurations, tool descriptions, and system prompts. It applies structural detectors (H1–H7) and Python pipeline detectors (P1–P2) to flag patterns such as vague instructions, missing constraints, and uncalibrated thresholds before runtime, and reports PASS, REVIEW, or FAIL verdicts. Scans are deterministic and make no LLM or network calls.

Findings identify bounded heuristic patterns. A clean scan does not establish runtime correctness, safety, or detector accuracy on external projects.