AI Agent 工程阅读清单

Five updates on model routing, sandbox snapshots, Cloud CLI MCP, Data Agent Kit, and sending production failures directly to coding agents.

Seven readings on browser authorization, task cost, isolated workspaces, MCP events, runtime safety, and multi-agent routing, with source limits intact.

Cloudflare cf, Kitesurf WebMCP, Claude Sonnet 5.5, and Forge: agent tool interfaces, browser tool calls, and schema-driven generation.

Seven updates on Codex prewarming and Guardian, remote sandboxes, MCP OAuth, secret stores, and agent software factory production data.

Five updates on tool and skill context budgets, context doctor, persistent scheduling, tool-loop limits, and fail-closed approvals.

Seven updates on token efficiency, Codex networking, Copilot Memory, local agent server security, self-practice, skills, and reproducibility.

Five updates on proactive safety intervention timing, state revision, TeamHarness, enterprise MCP permissions, and tool-result handling.

GitHub Copilot local sandboxing, Agensh, CliffCompaction, SWE-Serve, and enterprise OpenTelemetry support in the Copilot app.

Claude Opus 5.5 agent API compatibility, RRSI controls for harness overfitting, Harness-Zero behavior distillation, and Strands harness defaults.

LangGraph human-approval schemas, Deep Agents safeguards for parallel writes to the same file, and tighter OAuth boundaries in the MCP Ruby SDK.

OpenAI Agents SDK runtime-boundary fixes, stage-aware memory retrieval in AdaRepair-Mem, and static configuration checks with Agent Lint.

AI agent engineering readings for September 18 to 19, 2026, with key findings, engineering implications, suggested sections, and primary sources.