docs: enrich Context Compaction technical design detail
docs: update Sliding Prompt Caching detail in Tech Agent Design
docs: update agent design wiki with prompt cache, confirmations, and dynamic skills
docs: refresh agent design wiki
docs: sync agent design with runtime
docs: clarify context compaction threshold
docs: remove interview-specific wording
docs: keep simple and detailed runtime diagrams
docs: expand CLI runtime diagram
docs: add Chinese explanations for agent terms
docs: align wiki with current agent architecture
docs: refresh agent design wiki
docs: 重构 wiki 产品和 Agent 文档
docs: add TCP half-open timeout, LLM smart routing, exec_command exit sections
New sections in Agent Design wiki:
- Section 11: Stream chunk timeout (daemon thread + Queue pattern)
- Section 12: LLM smart routing (RoutingProvider + classify_turn)
- Section 13: exec_command and screen process exit handling
- Added model_start event to RuntimeEvent table
- Updated design principles table
Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
docs: add model compatibility section (GPT strict mode + DSv4 reasoning passback)
docs: rewrite all wiki pages for clarity and accuracy
- Rewrite all 10 pages in plain language, verify file paths against code
- Add new page: 11_Tech_Agent_Design (streaming, tools, context, memory, loop, providers, background tasks, skills)
- Fix outdated info: 5-layer funnel (not 4), correct module paths, Step2.5
- Update Home.md navigation
Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>