v2.0.9

What's New

• Fix: /search-chat no longer dumps the raw transcript, archive headers, or full script output into the user chat. Extracted content is now read internally and used to answer the user's question in the agent's own words, quoting only small, relevant snippets.
• Hardened skill instructions to prevent context-leak UX regressions and reinforce that archive data is input for analysis, not chat output.