Audit finds 89% of AI agent instructions are unverifiable promises

An analysis of eight agent-instruction collections (1,332 units, 17,611 instructions) found a median of only 11.1% of procedural instructions are checkable from the repository alone. Five of eight collections mandate zero output artifacts, and 43% of mandated artifacts across the other three have never existed in any commit. The tool's author warns that most instructions are "hopes with formatting" rather than enforceable rules.