Stop Returning Bare Linter Error Codes to Coding Agents: Opaque Error Codes vs. Actionable Fix-Hint Linters (-69.6% Input Tokens)

💡 AI Labo Benchmark Series — Experiment E17 (Feedback Sensors & Self-Correction) This post is part of our empirical benchmark series testing AI coding agent harness patterns inside network-isolated P