BestHub
Sign in
Architecture Development Notes
Author

Architecture Development Notes

Focused on architecture design, technology trend analysis, and practical development experience sharing.

109
Articles
0
Likes
876
Views
0
Comments
Recent Articles

Latest from Architecture Development Notes

100 recent articles max
Architecture Development Notes
Architecture Development Notes
Oct 21, 2026 · Artificial Intelligence

Claude Code 2.1.302: Fork Subagents Default On — Batch Tasks Lose Automatic Isolation

Claude Code 2.1.302 enables fork subagents by default in headless and editor entry points, breaking the assumption that batch subtasks run in isolated contexts; the article details trade-offs between fork (shared history, tools, model, cache, permissions) and custom subagents (fresh context, configurable tools/model), and guides choosing based on whether subtasks need shared context or independent isolation.

Agent SDKClaude Codebackground execution
0 likes · 26 min read
Claude Code 2.1.302: Fork Subagents Default On — Batch Tasks Lose Automatic Isolation
Architecture Development Notes
Architecture Development Notes
Oct 20, 2026 · Backend Development

ACP v2: Prompt Response Success Only Confirms Message Insertion, Not Turn Completion

The article explains how ACP v2 decouples message insertion from turn completion, making session/prompt responses only acknowledge insertion while deferring completion, errors, and cancellations to session-level state_update notifications, requiring clients to maintain separate tracking for submission receipt and session activity.

ACPError HandlingJSON-RPC
0 likes · 15 min read
ACP v2: Prompt Response Success Only Confirms Message Insertion, Not Turn Completion
Architecture Development Notes
Architecture Development Notes
Oct 18, 2026 · Artificial Intelligence

Claude Code Bug: Message Interruption Misreported as User Refusal

Claude Code 2.1.301 fixed a bug where tool calls interrupted by user messages during permission checks were incorrectly logged as user refusals, conflating explicit decisions with scheduling cancellations; the article argues for a distinct 'not_executed' status with cause and retryable fields to preserve accurate model behavior and audit metrics.

AI AgentClaude CodeTool Calling
0 likes · 9 min read
Claude Code Bug: Message Interruption Misreported as User Refusal
Architecture Development Notes
Architecture Development Notes
Oct 17, 2026 · Information Security

Sandbox Write to --plugin-dir Escapes to Host Execution: Claude Code 2.1.299 Analysis

Claude Code 2.1.299 fixed a vulnerability where sandboxed shell commands could write to the --plugin-dir directory, which the host then reloads and executes outside the sandbox, revealing that write protection must be dynamically derived from actual session-loaded code paths rather than static directory lists, and requiring cross-boundary testing that verifies write-to-execution chains.

Claude Codepath resolutionplugin directory
0 likes · 18 min read
Sandbox Write to --plugin-dir Escapes to Host Execution: Claude Code 2.1.299 Analysis
Architecture Development Notes
Architecture Development Notes
Oct 13, 2026 · Backend Development

Claude Code Agent State vs Status: Why 'Waiting for Answer' Loses Context After an Hour

The article explains how Claude Code's agent view JSON fields state, status, and waitingFor differ in semantics and lifecycle, revealing that 'blocked' sessions waiting for user responses lose their waitingFor field after about an hour when the supervisor kills the process, unlike permission-prompt waits which keep the process alive.

Claude CodeJSON APIagent monitoring
0 likes · 12 min read
Claude Code Agent State vs Status: Why 'Waiting for Answer' Loses Context After an Hour
Architecture Development Notes
Architecture Development Notes
Oct 12, 2026 · Backend Development

Claude Code Projects: Coordinator Blind to Thread History, Memory Becomes Source of Truth

Claude Code Projects coordinates work by spawning isolated cloud threads that report back results, but the coordinator sees only thread summaries — not full histories — forcing cross-thread decisions into persistent project memory, while a hard quota of 200 new threads per day replaces soft concurrency preferences.

AnthropicCLAUDE.mdClaude Code
0 likes · 12 min read
Claude Code Projects: Coordinator Blind to Thread History, Memory Becomes Source of Truth
Architecture Development Notes
Architecture Development Notes
Sep 11, 2026 · Artificial Intelligence

Control Inversion in Agent Tool Calling: Orchestration Loops Move Into Model-Generated Code

This article analyzes the shift from host-controlled to model-generated orchestration loops in AI agent tool calling, comparing Anthropic's API-level and LangChain's middleware approaches, examining cost control primitives, container lifecycle, debugging challenges, BFCL v4 evaluation data, and practical adoption criteria for programmatic tool calling.

AnthropicBFCL v4LangChain
0 likes · 17 min read
Control Inversion in Agent Tool Calling: Orchestration Loops Move Into Model-Generated Code
Architecture Development Notes
Architecture Development Notes
Sep 11, 2026 · Artificial Intelligence

Agent Delegation Identity: Building Auditable Chains for Tool Authorization

This article explains why AI agents need distinct delegated identities with auditable chains instead of shared secrets or user impersonation, detailing OAuth 2.0 Token Exchange (RFC 8693) act claims, nested delegation chains as audit evidence not authorization, MCP's resource indicator requirements, and practical patterns from Gravitee and Microsoft Entra Agent ID.

AI agentsDelegationMCP
0 likes · 17 min read
Agent Delegation Identity: Building Auditable Chains for Tool Authorization
Architecture Development Notes
Architecture Development Notes
Sep 8, 2026 · Artificial Intelligence

Rethinking Agent Composition: Single-Loop Skills vs. Sub-Agent Handoffs

This article analyzes why default multi-agent architectures leak state in long conversations, advocating for single-loop agents with dynamically loaded skills based on usage frequency, using Anthropic's commerce-agents reference implementation to illustrate caching-aware design, handoff vs. delegation distinctions, and evaluation strategies.

Agent architectureAnthropicLLM applications
0 likes · 10 min read
Rethinking Agent Composition: Single-Loop Skills vs. Sub-Agent Handoffs