What problem does it solve? Agents often quote stale or hallucinated progress numbers from chat memory instead of the actual disk state, causing the hub projection to drift from the real factory state. This Skill enforces reading disk truth before quoting any progress or hub state. ## Core Features & Use Cases - Agent truth bundle: Runs agent_truth_bundle_v1.py and factory_control_v1.py now at session start to load authoritative state. - Two-plane model: Separates the control plane (factory-now-v1.json, run-inbox-disk-truth-v1.json) from the projection plane (command-data.json, rebuilt via hub_projection_sync_v1). - Dual-pick law validation: Detects projection bugs by checking that live_pick equals queue_sa. - Use Case: Before replying with a status update, an agent runs the truth bundle, quotes the factory-now line, and flags any mismatch between goal-progress and run-inbox truth as a projection bug. ## Quick Start Ask the agent to run the truth bundle and quote the current factory-now line before reporting any hub progress.