What problem does it solve? Public chatbots drift out of sync with site content, producing stale or wrong answers about pricing, products, and developer paths. This Skill provides a structured audit ritual for the SourceA Brain chatbot so operators can verify knowledge freshness, run quality evals, and catch gaps before shipping worker or landing page changes. ## Core Features & Use Cases - SSOT manifest audit: Reads data/CHATBOT_KNOWLEDGE_MANIFEST.json to check sources, lanes, and stale flags before any Brain worker edit. - Distill/sync pipeline: Runs distill_www_to_brain_knowledge_v1.py, distill_docs_to_brain_knowledge_v1.py, and sync_brain_chat_knowledge_v1.py to rebuild the knowledge bundle from allowlisted sources. - Eval gates: Enforces P0/P1/P2/P3 eval buckets (90%/85% thresholds) via test_brain_chat_quality_v1.py, blocking deploys on failure. - Use Case: Before editing cloud/workers/sourcea-brain-chat-v1/ or updating landing page copy, load this Skill to run the Phase 1 audit ritual, refresh the knowledge bundle, and confirm the worker health endpoint reports the expected knowledge.bundle_version. ## Quick Start Ask the assistant to run the SourceA Brain chatbot audit ritual and check the knowledge manifest, gap report, and worker health before making any chatbot changes.
