Chat polish and runtime telemetry.

Enhanced Chat repo cards with owner avatar images, stable card reuse, tab render signatures, and scroll-preserving prompt logs so repository chats stay readable while updates stream in.

Standardized Dialogue references to Chat across dashboard sources, repo prompts, tabs, labels, and analytics context.

Hardened no-change follow-up routing by reading agent stdout from task logs, recognizing concrete no-op evidence, and avoiding escalation when the agent already cited repository evidence.

Accepted more non-fatal Hub snapshot refresh warnings when builds continue with existing snapshots and the admin snapshot key is unavailable.

Updated BigBrain runtime chat messages with local timestamps, assistant run-duration labels, approval stamps, and source badges grouped beside each bubble.

Merged live model and workspace usage events into the usage cache so token, cost, workspace status, and duration counters update without full query invalidation.

Filtered generic secret placeholders from missing-secret detection and limited content parsing to real missing-secret events.

Published a Railsmith agent-first blog post with its Codex workflow image and refreshed blog metadata.

Added AI citation and emotive brand audits with evidence, positioning recommendations, and artifact backlog for Molten.Bot content strategy.

Refined the README opening around Railsmith as an agent-readable engineering context maintenance loop, not only an AGENTS.md generator.